mjq11302010044 / RRPN

Arbitrary-Oriented Scene Text Detection via Rotation Proposals (TMM 2018)
Other
436 stars 141 forks source link

There is not enough memory on the GPU(8G) #16

Open wzhiyuan2016 opened 6 years ago

wzhiyuan2016 commented 6 years ago

Hi! Thank you for opening the source

My training configs set as follows


TRAIN': {'ASPECT_GROUPING': True, 'BATCH_SIZE': 64, 'BBOX_INSIDE_WEIGHTS': [1.0, 1.0, 1.0, 1.0], 'BBOX_NORMALIZE_MEANS': [0.0, 0.0, 0.0, 0.0], 'BBOX_NORMALIZE_STDS': [0.1, 0.1, 0.2, 0.2], 'BBOX_NORMALIZE_TARGETS': True, 'BBOX_NORMALIZE_TARGETS_PRECOMPUTED': True, 'BBOX_REG': True, 'BBOX_THRESH': 0.5, 'BG_THRESH_HI': 0.5, 'BG_THRESH_LO': 0.0, 'FG_FRACTION': 0.25, 'FG_THRESH': 0.5, 'GT_MARGIN': 1.4, 'HAS_RPN': True, 'IMS_PER_BATCH': 1, 'MAX_SIZE': 1000, 'PROPOSAL_METHOD': 'gt', 'RBBOX_INSIDE_WEIGHTS': [1.0, 1.0, 1.0, 1.0, 1.0], 'RBBOX_NORMALIZE_MEANS': [0.0, 0.0, 0.0, 0.0, 0.0], 'RBBOX_NORMALIZE_STDS': [0.1, 0.1, 0.2, 0.2, 1], 'RBBOX_NORMALIZE_TARGETS_PRECOMPUTED': True, 'RPN_BATCHSIZE': 256, 'RPN_BBOX_INSIDE_WEIGHTS': [1.0, 1.0, 1.0, 1.0], 'RPN_CLOBBER_POSITIVES': False, 'RPN_FG_FRACTION': 0.5, 'RPN_MIN_SIZE': 16, 'RPN_NEGATIVE_OVERLAP': 0.3, 'RPN_NMS_THRESH': 0.7, 'RPN_POSITIVE_OVERLAP': 0.7, 'RPN_POSITIVE_WEIGHT': -1.0, 'RPN_POST_NMS_TOP_N': 2000, 'RPN_PRE_NMS_TOP_N': 12000, 'RPN_RBBOX_INSIDE_WEIGHTS': [1.0, 1.0, 1.0, 1.0, 1.0], 'R_NEGATIVE_ANGLE_FILTER': 15, 'R_POSITIVE_ANGLE_FILTER': 15, 'SCALES': [600], 'SNAPSHOT_INFIX': '', 'SNAPSHOT_ITERS': 10000, 'USE_FLIPPED': False, 'USE_PREFETCH': False}, 'USE_GPU_NMS': True} and use ZF network


Images to use per minibatch __C.TRAIN.IMS_PER_BATCH = 1 I thought too much Anchors was generated,then I reduced the size, proportions, and angles of anchors, as follows:


def generate_anchors(base_size=16, ratios=[0.2,1], scales=2 ** np.arange(3,5), angle=[0.0,30.0]):


But there is always such a mistake Check failed: error == cudaSuccess (11 vs. 0) invalid argument

then, I set the size of A to 64,as follows: Minibatch size (number of regions of interest [ROIs]) A = __C.TRAIN.BATCH_SIZE = 64

always such a mistake as follows: Check failed: error == cudaSuccess (11 vs. 0) invalid argument As you released VGG16 model I can not download, I use the ZF.v2.caffemodel(imagenet_models) for fine-tuning training. Is it necessary to set what parameters? Or where did not notice. thank you!

mjq11302010044 commented 6 years ago

Do you compile the project with cudnn?

wzhiyuan2016 commented 6 years ago

yes, This is my running log file:


/home/x306/wzy/RRPN/tools/train_net.py --gpu 0 --solver /home/x306/wzy/RRPN/models/rrpn/ZF/faster_rcnn_end2end/solver.prototxt --weights /home/x306/wzy/py-faster-rcnn/data/imagenet_models/ZF.v2.caffemodel --imdb MSRA_TRAIN --iters 1000 --cfg /home/x306/wzy/RRPN/experiments/cfgs/faster_rcnn_end2end.yml Called with args: Namespace(cfg_file='/home/x306/wzy/RRPN/experiments/cfgs/faster_rcnn_end2end.yml', gpu_id=0, imdb_name='MSRA_TRAIN', max_iters=1000, pretrained_model='/home/x306/wzy/py-faster-rcnn/data/imagenet_models/ZF.v2.caffemodel', randomize=False, set_cfgs=None, solver='/home/x306/wzy/RRPN/models/rrpn/ZF/faster_rcnn_end2end/solver.prototxt') Using config: {'DATA_DIR': '/home/x306/wzy/RRPN/data', 'DEDUP_BOXES': 0.0625, 'EPS': 1e-14, 'EXP_DIR': 'faster_rcnn_end2end', 'GPU_ID': 0, 'IMG_PADDING': 0.25, 'MATLAB': 'matlab', 'MODELS_DIR': '/home/x306/wzy/RRPN/models/pascal_voc', 'PIXEL_MEANS': array([[[ 102.9801, 115.9465, 122.7717]]]), 'RNG_SEED': 3, 'ROOT_DIR': '/home/x306/wzy/RRPN', 'RRPN_MODELS_DIR': '/home/x306/wzy/RRPN/models/rrpn', 'TEST': {'ANGLE_GROUP': [], 'BBOX_REG': True, 'GT_MARGIN': 1.4, 'HAS_RPN': True, 'MAX_SIZE': 1700, 'NMS': 0.3, 'PROPOSAL_METHOD': 'selective_search', 'RATIO_GROUP': [0.2, 0.5, 1.0], 'RPN_MIN_SIZE': 16, 'RPN_NMS_THRESH': 0.7, 'RPN_POST_NMS_TOP_N': 300, 'RPN_PRE_NMS_TOP_N': 6000, 'SCALES': [1000], 'SCALE_GROUP': [], 'SVM': False}, 'TRAIN': {'ASPECT_GROUPING': True, 'BATCH_SIZE': 64, 'BBOX_INSIDE_WEIGHTS': [1.0, 1.0, 1.0, 1.0], 'BBOX_NORMALIZE_MEANS': [0.0, 0.0, 0.0, 0.0], 'BBOX_NORMALIZE_STDS': [0.1, 0.1, 0.2, 0.2], 'BBOX_NORMALIZE_TARGETS': True, 'BBOX_NORMALIZE_TARGETS_PRECOMPUTED': True, 'BBOX_REG': True, 'BBOX_THRESH': 0.5, 'BG_THRESH_HI': 0.5, 'BG_THRESH_LO': 0.0, 'FG_FRACTION': 0.25, 'FG_THRESH': 0.5, 'GT_MARGIN': 1.4, 'HAS_RPN': True, 'IMS_PER_BATCH': 1, 'MAX_SIZE': 1000, 'PROPOSAL_METHOD': 'gt', 'RBBOX_INSIDE_WEIGHTS': [1.0, 1.0, 1.0, 1.0, 1.0], 'RBBOX_NORMALIZE_MEANS': [0.0, 0.0, 0.0, 0.0, 0.0], 'RBBOX_NORMALIZE_STDS': [0.1, 0.1, 0.2, 0.2, 1], 'RBBOX_NORMALIZE_TARGETS_PRECOMPUTED': True, 'RPN_BATCHSIZE': 256, 'RPN_BBOX_INSIDE_WEIGHTS': [1.0, 1.0, 1.0, 1.0], 'RPN_CLOBBER_POSITIVES': False, 'RPN_FG_FRACTION': 0.5, 'RPN_MIN_SIZE': 16, 'RPN_NEGATIVE_OVERLAP': 0.3, 'RPN_NMS_THRESH': 0.7, 'RPN_POSITIVE_OVERLAP': 0.7, 'RPN_POSITIVE_WEIGHT': -1.0, 'RPN_POST_NMS_TOP_N': 2000, 'RPN_PRE_NMS_TOP_N': 12000, 'RPN_RBBOX_INSIDE_WEIGHTS': [1.0, 1.0, 1.0, 1.0, 1.0], 'R_NEGATIVE_ANGLE_FILTER': 15, 'R_POSITIVE_ANGLE_FILTER': 15, 'SCALES': [600], 'SNAPSHOT_INFIX': '', 'SNAPSHOT_ITERS': 10000, 'USE_FLIPPED': False, 'USE_PREFETCH': False}, 'USE_GPU_NMS': True} MSRA_TRAIN 1 roidb entries Output will be saved to . Filtered 0 roidb entries: 1 -> 1 Computing bounding-box regression targets... rtrain中的SolverWrapper类Computing bounding-box regression targets... 调用了rbbox中的 add_rbbox_regression_target函数 调用了rbbox_transform中的rbbox_transform bbox target means: [[ 0. 0. 0. 0. 0.] [ 0. 0. 0. 0. 0.]] [ 0. 0. 0. 0. 0.] bbox target stdevs: [[ 0.1 0.1 0.2 0.2 1. ] [ 0.1 0.1 0.2 0.2 1. ]] [ 0.1 0.1 0.2 0.2 1. ] Normalizing targets done RoiDataLayer: name_to_top: {'gt_boxes': 2, 'data': 0, 'im_info': 1} 调用了rlayer中的 rRoIDataLaye Setup完成 WARNING: Logging before InitGoogleLogging() is written to STDERR I0130 16:30:48.655997 18755 solver.cpp:48] Initializing solver from parameters: train_net: "/home/x306/wzy/RRPN/models/rrpn/ZF/faster_rcnn_end2end/train.prototxt" base_lr: 0.001 display: 20 lr_policy: "step" gamma: 0.1 momentum: 0.9 weight_decay: 0.0005 stepsize: 50000 snapshot: 0 snapshot_prefix: "zf_faster_rcnn" average_loss: 100 iter_size: 4 I0130 16:30:48.656019 18755 solver.cpp:81] Creating training net from train_net file: /home/x306/wzy/RRPN/models/rrpn/ZF/faster_rcnn_end2end/train.prototxt I0130 16:30:48.656664 18755 net.cpp:49] Initializing net from parameters: name: "ZF" I0130 16:30:48.656795 18755 layer_factory.hpp:77] Creating layer input-data I0130 16:30:48.657094 18755 net.cpp:106] Creating Layer input-data I0130 16:30:48.657101 18755 net.cpp:411] input-data -> data I0130 16:30:48.657109 18755 net.cpp:411] input-data -> im_info I0130 16:30:48.657112 18755 net.cpp:411] input-data -> gt_boxes I0130 16:30:48.658023 18755 net.cpp:150] Setting up input-data I0130 16:30:48.658031 18755 net.cpp:157] Top shape: 1 3 600 1000 (1800000) I0130 16:30:48.658035 18755 net.cpp:157] Top shape: 1 3 (3) I0130 16:30:48.658038 18755 net.cpp:157] Top shape: 1 5 (5) I0130 16:30:48.658040 18755 net.cpp:165] Memory required for data: 7200032 I0130 16:30:48.658042 18755 layer_factory.hpp:77] Creating layer data_input-data_0_split I0130 16:30:48.658047 18755 net.cpp:106] Creating Layer data_input-data_0_split I0130 16:30:48.658049 18755 net.cpp:454] data_input-data_0_split <- data I0130 16:30:48.658053 18755 net.cpp:411] data_input-data_0_split -> data_input-data_0_split_0 I0130 16:30:48.658058 18755 net.cpp:411] data_input-data_0_split -> data_input-data_0_split_1 I0130 16:30:48.658077 18755 net.cpp:150] Setting up data_input-data_0_split I0130 16:30:48.658082 18755 net.cpp:157] Top shape: 1 3 600 1000 (1800000) I0130 16:30:48.658084 18755 net.cpp:157] Top shape: 1 3 600 1000 (1800000) I0130 16:30:48.658087 18755 net.cpp:165] Memory required for data: 21600032 I0130 16:30:48.658087 18755 layer_factory.hpp:77] Creating layer im_info_input-data_1_split I0130 16:30:48.658092 18755 net.cpp:106] Creating Layer im_info_input-data_1_split I0130 16:30:48.658093 18755 net.cpp:454] im_info_input-data_1_split <- im_info I0130 16:30:48.658097 18755 net.cpp:411] im_info_input-data_1_split -> im_info_input-data_1_split_0 I0130 16:30:48.658099 18755 net.cpp:411] im_info_input-data_1_split -> im_info_input-data_1_split_1 I0130 16:30:48.658102 18755 net.cpp:411] im_info_input-data_1_split -> im_info_input-data_1_split_2 I0130 16:30:48.658124 18755 net.cpp:150] Setting up im_info_input-data_1_split I0130 16:30:48.658128 18755 net.cpp:157] Top shape: 1 3 (3) I0130 16:30:48.658129 18755 net.cpp:157] Top shape: 1 3 (3) I0130 16:30:48.658133 18755 net.cpp:157] Top shape: 1 3 (3) I0130 16:30:48.658133 18755 net.cpp:165] Memory required for data: 21600068 I0130 16:30:48.658135 18755 layer_factory.hpp:77] Creating layer gt_boxes_input-data_2_split I0130 16:30:48.658138 18755 net.cpp:106] Creating Layer gt_boxes_input-data_2_split I0130 16:30:48.658140 18755 net.cpp:454] gt_boxes_input-data_2_split <- gt_boxes I0130 16:30:48.658143 18755 net.cpp:411] gt_boxes_input-data_2_split -> gt_boxes_input-data_2_split_0 I0130 16:30:48.658148 18755 net.cpp:411] gt_boxes_input-data_2_split -> gt_boxes_input-data_2_split_1 I0130 16:30:48.658162 18755 net.cpp:150] Setting up gt_boxes_input-data_2_split I0130 16:30:48.658165 18755 net.cpp:157] Top shape: 1 5 (5) I0130 16:30:48.658167 18755 net.cpp:157] Top shape: 1 5 (5) I0130 16:30:48.658169 18755 net.cpp:165] Memory required for data: 21600108 I0130 16:30:48.658171 18755 layer_factory.hpp:77] Creating layer conv1 I0130 16:30:48.658176 18755 net.cpp:106] Creating Layer conv1 I0130 16:30:48.658179 18755 net.cpp:454] conv1 <- data_input-data_0_split_0 I0130 16:30:48.658182 18755 net.cpp:411] conv1 -> conv1 I0130 16:30:48.850282 18755 net.cpp:150] Setting up conv1 I0130 16:30:48.850301 18755 net.cpp:157] Top shape: 1 96 300 500 (14400000) I0130 16:30:48.850303 18755 net.cpp:165] Memory required for data: 79200108 I0130 16:30:48.850316 18755 layer_factory.hpp:77] Creating layer relu1 I0130 16:30:48.850327 18755 net.cpp:106] Creating Layer relu1 I0130 16:30:48.850330 18755 net.cpp:454] relu1 <- conv1 I0130 16:30:48.850337 18755 net.cpp:397] relu1 -> conv1 (in-place) I0130 16:30:48.850471 18755 net.cpp:150] Setting up relu1 I0130 16:30:48.850477 18755 net.cpp:157] Top shape: 1 96 300 500 (14400000) I0130 16:30:48.850481 18755 net.cpp:165] Memory required for data: 136800108 I0130 16:30:48.850482 18755 layer_factory.hpp:77] Creating layer norm1 I0130 16:30:48.850489 18755 net.cpp:106] Creating Layer norm1 I0130 16:30:48.850492 18755 net.cpp:454] norm1 <- conv1 I0130 16:30:48.850495 18755 net.cpp:411] norm1 -> norm1 I0130 16:30:48.850576 18755 net.cpp:150] Setting up norm1 I0130 16:30:48.850582 18755 net.cpp:157] Top shape: 1 96 300 500 (14400000) I0130 16:30:48.850584 18755 net.cpp:165] Memory required for data: 194400108 I0130 16:30:48.850586 18755 layer_factory.hpp:77] Creating layer pool1 I0130 16:30:48.850591 18755 net.cpp:106] Creating Layer pool1 I0130 16:30:48.850594 18755 net.cpp:454] pool1 <- norm1 I0130 16:30:48.850597 18755 net.cpp:411] pool1 -> pool1 I0130 16:30:48.850616 18755 net.cpp:150] Setting up pool1 I0130 16:30:48.850620 18755 net.cpp:157] Top shape: 1 96 151 251 (3638496) I0130 16:30:48.850621 18755 net.cpp:165] Memory required for data: 208954092 I0130 16:30:48.850625 18755 layer_factory.hpp:77] Creating layer conv2 I0130 16:30:48.850632 18755 net.cpp:106] Creating Layer conv2 I0130 16:30:48.850634 18755 net.cpp:454] conv2 <- pool1 I0130 16:30:48.850637 18755 net.cpp:411] conv2 -> conv2 I0130 16:30:48.852658 18755 net.cpp:150] Setting up conv2 I0130 16:30:48.852669 18755 net.cpp:157] Top shape: 1 256 76 126 (2451456) I0130 16:30:48.852671 18755 net.cpp:165] Memory required for data: 218759916 I0130 16:30:48.852679 18755 layer_factory.hpp:77] Creating layer relu2 I0130 16:30:48.852685 18755 net.cpp:106] Creating Layer relu2 I0130 16:30:48.852687 18755 net.cpp:454] relu2 <- conv2 I0130 16:30:48.852691 18755 net.cpp:397] relu2 -> conv2 (in-place) I0130 16:30:48.852798 18755 net.cpp:150] Setting up relu2 I0130 16:30:48.852804 18755 net.cpp:157] Top shape: 1 256 76 126 (2451456) I0130 16:30:48.852807 18755 net.cpp:165] Memory required for data: 228565740 I0130 16:30:48.852809 18755 layer_factory.hpp:77] Creating layer norm2 I0130 16:30:48.852815 18755 net.cpp:106] Creating Layer norm2 I0130 16:30:48.852818 18755 net.cpp:454] norm2 <- conv2 I0130 16:30:48.852824 18755 net.cpp:411] norm2 -> norm2 I0130 16:30:48.852893 18755 net.cpp:150] Setting up norm2 I0130 16:30:48.852898 18755 net.cpp:157] Top shape: 1 256 76 126 (2451456) I0130 16:30:48.852901 18755 net.cpp:165] Memory required for data: 238371564 I0130 16:30:48.852903 18755 layer_factory.hpp:77] Creating layer pool2 I0130 16:30:48.852906 18755 net.cpp:106] Creating Layer pool2 I0130 16:30:48.852910 18755 net.cpp:454] pool2 <- norm2 I0130 16:30:48.852912 18755 net.cpp:411] pool2 -> pool2 I0130 16:30:48.852931 18755 net.cpp:150] Setting up pool2 I0130 16:30:48.852934 18755 net.cpp:157] Top shape: 1 256 39 64 (638976) I0130 16:30:48.852936 18755 net.cpp:165] Memory required for data: 240927468 I0130 16:30:48.852938 18755 layer_factory.hpp:77] Creating layer conv3 I0130 16:30:48.852944 18755 net.cpp:106] Creating Layer conv3 I0130 16:30:48.852946 18755 net.cpp:454] conv3 <- pool2 I0130 16:30:48.852949 18755 net.cpp:411] conv3 -> conv3 I0130 16:30:48.856401 18755 net.cpp:150] Setting up conv3 I0130 16:30:48.856418 18755 net.cpp:157] Top shape: 1 384 39 64 (958464) I0130 16:30:48.856421 18755 net.cpp:165] Memory required for data: 244761324 I0130 16:30:48.856431 18755 layer_factory.hpp:77] Creating layer relu3 I0130 16:30:48.856438 18755 net.cpp:106] Creating Layer relu3 I0130 16:30:48.856441 18755 net.cpp:454] relu3 <- conv3 I0130 16:30:48.856446 18755 net.cpp:397] relu3 -> conv3 (in-place) I0130 16:30:48.856884 18755 net.cpp:150] Setting up relu3 I0130 16:30:48.856891 18755 net.cpp:157] Top shape: 1 384 39 64 (958464) I0130 16:30:48.856894 18755 net.cpp:165] Memory required for data: 248595180 I0130 16:30:48.856896 18755 layer_factory.hpp:77] Creating layer conv4 I0130 16:30:48.856904 18755 net.cpp:106] Creating Layer conv4 I0130 16:30:48.856906 18755 net.cpp:454] conv4 <- conv3 I0130 16:30:48.856911 18755 net.cpp:411] conv4 -> conv4 I0130 16:30:48.859576 18755 net.cpp:150] Setting up conv4 I0130 16:30:48.859592 18755 net.cpp:157] Top shape: 1 384 39 64 (958464) I0130 16:30:48.859594 18755 net.cpp:165] Memory required for data: 252429036 I0130 16:30:48.859601 18755 layer_factory.hpp:77] Creating layer relu4 I0130 16:30:48.859607 18755 net.cpp:106] Creating Layer relu4 I0130 16:30:48.859611 18755 net.cpp:454] relu4 <- conv4 I0130 16:30:48.859614 18755 net.cpp:397] relu4 -> conv4 (in-place) I0130 16:30:48.859753 18755 net.cpp:150] Setting up relu4 I0130 16:30:48.859760 18755 net.cpp:157] Top shape: 1 384 39 64 (958464) I0130 16:30:48.859761 18755 net.cpp:165] Memory required for data: 256262892 I0130 16:30:48.859763 18755 layer_factory.hpp:77] Creating layer conv5 I0130 16:30:48.859772 18755 net.cpp:106] Creating Layer conv5 I0130 16:30:48.859774 18755 net.cpp:454] conv5 <- conv4 I0130 16:30:48.859779 18755 net.cpp:411] conv5 -> conv5 I0130 16:30:48.862138 18755 net.cpp:150] Setting up conv5 I0130 16:30:48.862149 18755 net.cpp:157] Top shape: 1 256 39 64 (638976) I0130 16:30:48.862152 18755 net.cpp:165] Memory required for data: 258818796 I0130 16:30:48.862160 18755 layer_factory.hpp:77] Creating layer relu5 I0130 16:30:48.862165 18755 net.cpp:106] Creating Layer relu5 I0130 16:30:48.862169 18755 net.cpp:454] relu5 <- conv5 I0130 16:30:48.862172 18755 net.cpp:397] relu5 -> conv5 (in-place) I0130 16:30:48.862288 18755 net.cpp:150] Setting up relu5 I0130 16:30:48.862294 18755 net.cpp:157] Top shape: 1 256 39 64 (638976) I0130 16:30:48.862296 18755 net.cpp:165] Memory required for data: 261374700 I0130 16:30:48.862298 18755 layer_factory.hpp:77] Creating layer conv5_relu5_0_split I0130 16:30:48.862303 18755 net.cpp:106] Creating Layer conv5_relu5_0_split I0130 16:30:48.862305 18755 net.cpp:454] conv5_relu5_0_split <- conv5 I0130 16:30:48.862308 18755 net.cpp:411] conv5_relu5_0_split -> conv5_relu5_0_split_0 I0130 16:30:48.862313 18755 net.cpp:411] conv5_relu5_0_split -> conv5_relu5_0_split_1 I0130 16:30:48.862339 18755 net.cpp:150] Setting up conv5_relu5_0_split I0130 16:30:48.862342 18755 net.cpp:157] Top shape: 1 256 39 64 (638976) I0130 16:30:48.862344 18755 net.cpp:157] Top shape: 1 256 39 64 (638976) I0130 16:30:48.862346 18755 net.cpp:165] Memory required for data: 266486508 I0130 16:30:48.862349 18755 layer_factory.hpp:77] Creating layer rpn_conv/3x3 I0130 16:30:48.862355 18755 net.cpp:106] Creating Layer rpn_conv/3x3 I0130 16:30:48.862357 18755 net.cpp:454] rpn_conv/3x3 <- conv5_relu5_0_split_0 I0130 16:30:48.862361 18755 net.cpp:411] rpn_conv/3x3 -> rpn/output I0130 16:30:48.877441 18755 net.cpp:150] Setting up rpn_conv/3x3 I0130 16:30:48.877470 18755 net.cpp:157] Top shape: 1 256 39 64 (638976) I0130 16:30:48.877471 18755 net.cpp:165] Memory required for data: 269042412 I0130 16:30:48.877478 18755 layer_factory.hpp:77] Creating layer rpn_relu/3x3 I0130 16:30:48.877486 18755 net.cpp:106] Creating Layer rpn_relu/3x3 I0130 16:30:48.877490 18755 net.cpp:454] rpn_relu/3x3 <- rpn/output I0130 16:30:48.877495 18755 net.cpp:397] rpn_relu/3x3 -> rpn/output (in-place) I0130 16:30:48.877609 18755 net.cpp:150] Setting up rpn_relu/3x3 I0130 16:30:48.877614 18755 net.cpp:157] Top shape: 1 256 39 64 (638976) I0130 16:30:48.877616 18755 net.cpp:165] Memory required for data: 271598316 I0130 16:30:48.877619 18755 layer_factory.hpp:77] Creating layer rpn/output_rpn_relu/3x3_0_split I0130 16:30:48.877622 18755 net.cpp:106] Creating Layer rpn/output_rpn_relu/3x3_0_split I0130 16:30:48.877625 18755 net.cpp:454] rpn/output_rpn_relu/3x3_0_split <- rpn/output I0130 16:30:48.877629 18755 net.cpp:411] rpn/output_rpn_relu/3x3_0_split -> rpn/output_rpn_relu/3x3_0_split_0 I0130 16:30:48.877634 18755 net.cpp:411] rpn/output_rpn_relu/3x3_0_split -> rpn/output_rpn_relu/3x3_0_split_1 I0130 16:30:48.877658 18755 net.cpp:150] Setting up rpn/output_rpn_relu/3x3_0_split I0130 16:30:48.877661 18755 net.cpp:157] Top shape: 1 256 39 64 (638976) I0130 16:30:48.877665 18755 net.cpp:157] Top shape: 1 256 39 64 (638976) I0130 16:30:48.877666 18755 net.cpp:165] Memory required for data: 276710124 I0130 16:30:48.877668 18755 layer_factory.hpp:77] Creating layer rpn_cls_score I0130 16:30:48.877676 18755 net.cpp:106] Creating Layer rpn_cls_score I0130 16:30:48.877678 18755 net.cpp:454] rpn_cls_score <- rpn/output_rpn_relu/3x3_0_split_0 I0130 16:30:48.877682 18755 net.cpp:411] rpn_cls_score -> rpn_cls_score I0130 16:30:48.878798 18755 net.cpp:150] Setting up rpn_cls_score I0130 16:30:48.878810 18755 net.cpp:157] Top shape: 1 36 39 64 (89856) I0130 16:30:48.878813 18755 net.cpp:165] Memory required for data: 277069548 I0130 16:30:48.878818 18755 layer_factory.hpp:77] Creating layer rpn_cls_score_rpn_cls_score_0_split I0130 16:30:48.878826 18755 net.cpp:106] Creating Layer rpn_cls_score_rpn_cls_score_0_split I0130 16:30:48.878829 18755 net.cpp:454] rpn_cls_score_rpn_cls_score_0_split <- rpn_cls_score I0130 16:30:48.878834 18755 net.cpp:411] rpn_cls_score_rpn_cls_score_0_split -> rpn_cls_score_rpn_cls_score_0_split_0 I0130 16:30:48.878839 18755 net.cpp:411] rpn_cls_score_rpn_cls_score_0_split -> rpn_cls_score_rpn_cls_score_0_split_1 I0130 16:30:48.878865 18755 net.cpp:150] Setting up rpn_cls_score_rpn_cls_score_0_split I0130 16:30:48.878867 18755 net.cpp:157] Top shape: 1 36 39 64 (89856) I0130 16:30:48.878870 18755 net.cpp:157] Top shape: 1 36 39 64 (89856) I0130 16:30:48.878872 18755 net.cpp:165] Memory required for data: 277788396 I0130 16:30:48.878875 18755 layer_factory.hpp:77] Creating layer rpn_bbox_pred I0130 16:30:48.878883 18755 net.cpp:106] Creating Layer rpn_bbox_pred I0130 16:30:48.878885 18755 net.cpp:454] rpn_bbox_pred <- rpn/output_rpn_relu/3x3_0_split_1 I0130 16:30:48.878890 18755 net.cpp:411] rpn_bbox_pred -> rpn_bbox_pred I0130 16:30:48.882532 18755 net.cpp:150] Setting up rpn_bbox_pred I0130 16:30:48.882593 18755 net.cpp:157] Top shape: 1 90 39 64 (224640) I0130 16:30:48.882598 18755 net.cpp:165] Memory required for data: 278686956 I0130 16:30:48.882612 18755 layer_factory.hpp:77] Creating layer rpn_bbox_pred_rpn_bbox_pred_0_split I0130 16:30:48.882623 18755 net.cpp:106] Creating Layer rpn_bbox_pred_rpn_bbox_pred_0_split I0130 16:30:48.882630 18755 net.cpp:454] rpn_bbox_pred_rpn_bbox_pred_0_split <- rpn_bbox_pred I0130 16:30:48.882639 18755 net.cpp:411] rpn_bbox_pred_rpn_bbox_pred_0_split -> rpn_bbox_pred_rpn_bbox_pred_0_split_0 I0130 16:30:48.882649 18755 net.cpp:411] rpn_bbox_pred_rpn_bbox_pred_0_split -> rpn_bbox_pred_rpn_bbox_pred_0_split_1 I0130 16:30:48.882685 18755 net.cpp:150] Setting up rpn_bbox_pred_rpn_bbox_pred_0_split I0130 16:30:48.882690 18755 net.cpp:157] Top shape: 1 90 39 64 (224640) I0130 16:30:48.882694 18755 net.cpp:157] Top shape: 1 90 39 64 (224640) I0130 16:30:48.882695 18755 net.cpp:165] Memory required for data: 280484076 I0130 16:30:48.882697 18755 layer_factory.hpp:77] Creating layer rpn_cls_score_reshape I0130 16:30:48.882704 18755 net.cpp:106] Creating Layer rpn_cls_score_reshape I0130 16:30:48.882706 18755 net.cpp:454] rpn_cls_score_reshape <- rpn_cls_score_rpn_cls_score_0_split_0 I0130 16:30:48.882711 18755 net.cpp:411] rpn_cls_score_reshape -> rpn_cls_score_reshape I0130 16:30:48.882727 18755 net.cpp:150] Setting up rpn_cls_score_reshape I0130 16:30:48.882731 18755 net.cpp:157] Top shape: 1 2 702 64 (89856) I0130 16:30:48.882733 18755 net.cpp:165] Memory required for data: 280843500 I0130 16:30:48.882735 18755 layer_factory.hpp:77] Creating layer rpn_cls_score_reshape_rpn_cls_score_reshape_0_split I0130 16:30:48.882738 18755 net.cpp:106] Creating Layer rpn_cls_score_reshape_rpn_cls_score_reshape_0_split I0130 16:30:48.882741 18755 net.cpp:454] rpn_cls_score_reshape_rpn_cls_score_reshape_0_split <- rpn_cls_score_reshape I0130 16:30:48.882745 18755 net.cpp:411] rpn_cls_score_reshape_rpn_cls_score_reshape_0_split -> rpn_cls_score_reshape_rpn_cls_score_reshape_0_split_0 I0130 16:30:48.882748 18755 net.cpp:411] rpn_cls_score_reshape_rpn_cls_score_reshape_0_split -> rpn_cls_score_reshape_rpn_cls_score_reshape_0_split_1 I0130 16:30:48.882771 18755 net.cpp:150] Setting up rpn_cls_score_reshape_rpn_cls_score_reshape_0_split I0130 16:30:48.882774 18755 net.cpp:157] Top shape: 1 2 702 64 (89856) I0130 16:30:48.882777 18755 net.cpp:157] Top shape: 1 2 702 64 (89856) I0130 16:30:48.882779 18755 net.cpp:165] Memory required for data: 281562348 I0130 16:30:48.882781 18755 layer_factory.hpp:77] Creating layer rpn-data I0130 16:30:48.883280 18755 net.cpp:106] Creating Layer rpn-data I0130 16:30:48.883287 18755 net.cpp:454] rpn-data <- rpn_cls_score_rpn_cls_score_0_split_1 I0130 16:30:48.883291 18755 net.cpp:454] rpn-data <- gt_boxes_input-data_2_split_0 I0130 16:30:48.883294 18755 net.cpp:454] rpn-data <- im_info_input-data_1_split_0 I0130 16:30:48.883298 18755 net.cpp:454] rpn-data <- data_input-data_0_split_1 I0130 16:30:48.883301 18755 net.cpp:411] rpn-data -> rpn_labels I0130 16:30:48.883316 18755 net.cpp:411] rpn-data -> rpn_bbox_targets I0130 16:30:48.883321 18755 net.cpp:411] rpn-data -> rpn_bbox_inside_weights I0130 16:30:48.883324 18755 net.cpp:411] rpn-data -> rpn_bbox_outside_weights I0130 16:30:48.884225 18755 net.cpp:150] Setting up rpn-data I0130 16:30:48.884250 18755 net.cpp:157] Top shape: 1 1 702 64 (44928) I0130 16:30:48.884256 18755 net.cpp:157] Top shape: 1 90 39 64 (224640) I0130 16:30:48.884261 18755 net.cpp:157] Top shape: 1 90 39 64 (224640) I0130 16:30:48.884265 18755 net.cpp:157] Top shape: 1 90 39 64 (224640) I0130 16:30:48.884268 18755 net.cpp:165] Memory required for data: 284437740 I0130 16:30:48.884270 18755 layer_factory.hpp:77] Creating layer rpn_loss_cls I0130 16:30:48.884276 18755 net.cpp:106] Creating Layer rpn_loss_cls I0130 16:30:48.884279 18755 net.cpp:454] rpn_loss_cls <- rpn_cls_score_reshape_rpn_cls_score_reshape_0_split_0 I0130 16:30:48.884284 18755 net.cpp:454] rpn_loss_cls <- rpn_labels I0130 16:30:48.884289 18755 net.cpp:411] rpn_loss_cls -> rpn_cls_loss I0130 16:30:48.884294 18755 layer_factory.hpp:77] Creating layer rpn_loss_cls I0130 16:30:48.884610 18755 net.cpp:150] Setting up rpn_loss_cls I0130 16:30:48.884618 18755 net.cpp:157] Top shape: (1) I0130 16:30:48.884619 18755 net.cpp:160] with loss weight 1 I0130 16:30:48.884629 18755 net.cpp:165] Memory required for data: 284437744 I0130 16:30:48.884631 18755 layer_factory.hpp:77] Creating layer rpn_loss_bbox I0130 16:30:48.884637 18755 net.cpp:106] Creating Layer rpn_loss_bbox I0130 16:30:48.884649 18755 net.cpp:454] rpn_loss_bbox <- rpn_bbox_pred_rpn_bbox_pred_0_split_0 I0130 16:30:48.884654 18755 net.cpp:454] rpn_loss_bbox <- rpn_bbox_targets I0130 16:30:48.884656 18755 net.cpp:454] rpn_loss_bbox <- rpn_bbox_inside_weights I0130 16:30:48.884668 18755 net.cpp:454] rpn_loss_bbox <- rpn_bbox_outside_weights I0130 16:30:48.884672 18755 net.cpp:411] rpn_loss_bbox -> rpn_loss_bbox I0130 16:30:48.887003 18755 net.cpp:150] Setting up rpn_loss_bbox I0130 16:30:48.887028 18755 net.cpp:157] Top shape: (1) I0130 16:30:48.887032 18755 net.cpp:160] with loss weight 1 I0130 16:30:48.887040 18755 net.cpp:165] Memory required for data: 284437748 I0130 16:30:48.887045 18755 layer_factory.hpp:77] Creating layer rpn_cls_prob I0130 16:30:48.887055 18755 net.cpp:106] Creating Layer rpn_cls_prob I0130 16:30:48.887060 18755 net.cpp:454] rpn_cls_prob <- rpn_cls_score_reshape_rpn_cls_score_reshape_0_split_1 I0130 16:30:48.887068 18755 net.cpp:411] rpn_cls_prob -> rpn_cls_prob I0130 16:30:48.887296 18755 net.cpp:150] Setting up rpn_cls_prob I0130 16:30:48.887303 18755 net.cpp:157] Top shape: 1 2 702 64 (89856) I0130 16:30:48.887306 18755 net.cpp:165] Memory required for data: 284797172 I0130 16:30:48.887308 18755 layer_factory.hpp:77] Creating layer rpn_cls_prob_reshape I0130 16:30:48.887320 18755 net.cpp:106] Creating Layer rpn_cls_prob_reshape I0130 16:30:48.887323 18755 net.cpp:454] rpn_cls_prob_reshape <- rpn_cls_prob I0130 16:30:48.887327 18755 net.cpp:411] rpn_cls_prob_reshape -> rpn_cls_prob_reshape I0130 16:30:48.887348 18755 net.cpp:150] Setting up rpn_cls_prob_reshape I0130 16:30:48.887352 18755 net.cpp:157] Top shape: 1 36 39 64 (89856) I0130 16:30:48.887354 18755 net.cpp:165] Memory required for data: 285156596 I0130 16:30:48.887356 18755 layer_factory.hpp:77] Creating layer proposal I0130 16:30:48.887778 18755 net.cpp:106] Creating Layer proposal I0130 16:30:48.887786 18755 net.cpp:454] proposal <- rpn_cls_prob_reshape I0130 16:30:48.887790 18755 net.cpp:454] proposal <- rpn_bbox_pred_rpn_bbox_pred_0_split_1 I0130 16:30:48.887794 18755 net.cpp:454] proposal <- im_info_input-data_1_split_1 I0130 16:30:48.887797 18755 net.cpp:411] proposal -> rpn_rois I0130 16:30:48.889271 18755 net.cpp:150] Setting up proposal I0130 16:30:48.889283 18755 net.cpp:157] Top shape: 1 6 (6) I0130 16:30:48.889286 18755 net.cpp:165] Memory required for data: 285156620 I0130 16:30:48.889289 18755 layer_factory.hpp:77] Creating layer roi-data I0130 16:30:48.889443 18755 net.cpp:106] Creating Layer roi-data I0130 16:30:48.889451 18755 net.cpp:454] roi-data <- rpn_rois I0130 16:30:48.889456 18755 net.cpp:454] roi-data <- gt_boxes_input-data_2_split_1 I0130 16:30:48.889459 18755 net.cpp:411] roi-data -> rois I0130 16:30:48.889466 18755 net.cpp:411] roi-data -> labels I0130 16:30:48.889470 18755 net.cpp:411] roi-data -> bbox_targets I0130 16:30:48.889474 18755 net.cpp:411] roi-data -> bbox_inside_weights I0130 16:30:48.889478 18755 net.cpp:411] roi-data -> bbox_outside_weights I0130 16:30:48.889778 18755 net.cpp:150] Setting up roi-data I0130 16:30:48.889786 18755 net.cpp:157] Top shape: 1 6 (6) I0130 16:30:48.889789 18755 net.cpp:157] Top shape: 1 1 (1) I0130 16:30:48.889791 18755 net.cpp:157] Top shape: 1 10 (10) I0130 16:30:48.889794 18755 net.cpp:157] Top shape: 1 10 (10) I0130 16:30:48.889797 18755 net.cpp:157] Top shape: 1 10 (10) I0130 16:30:48.889799 18755 net.cpp:165] Memory required for data: 285156768 I0130 16:30:48.889801 18755 layer_factory.hpp:77] Creating layer roi_pool_conv5 I0130 16:30:48.889807 18755 net.cpp:106] Creating Layer roi_pool_conv5 I0130 16:30:48.889811 18755 net.cpp:454] roi_pool_conv5 <- conv5_relu5_0_split_1 I0130 16:30:48.889814 18755 net.cpp:454] roi_pool_conv5 <- rois I0130 16:30:48.889817 18755 net.cpp:454] roi_pool_conv5 <- im_info_input-data_1_split_2 I0130 16:30:48.889822 18755 net.cpp:411] roi_pool_conv5 -> roi_pool_conv5 I0130 16:30:48.889827 18755 rotate_roi_pooling_layer.cpp:24] Spatial scale: 0.0625 I0130 16:30:48.889858 18755 net.cpp:150] Setting up roi_pool_conv5 I0130 16:30:48.889861 18755 net.cpp:157] Top shape: 1 256 6 6 (9216) I0130 16:30:48.889864 18755 net.cpp:165] Memory required for data: 285193632 I0130 16:30:48.889866 18755 layer_factory.hpp:77] Creating layer fc6 I0130 16:30:48.889871 18755 net.cpp:106] Creating Layer fc6 I0130 16:30:48.889874 18755 net.cpp:454] fc6 <- roi_pool_conv5 I0130 16:30:48.889878 18755 net.cpp:411] fc6 -> fc6 3.ProposalTargetLayer setup 第三次产生Anchors完成 3.ProposalTargetLayer 调用了reshape 第三次产生Anchors I0130 16:30:48.976393 18755 net.cpp:150] Setting up fc6 I0130 16:30:48.976436 18755 net.cpp:157] Top shape: 1 4096 (4096) I0130 16:30:48.976444 18755 net.cpp:165] Memory required for data: 285210016 I0130 16:30:48.976469 18755 layer_factory.hpp:77] Creating layer relu6 I0130 16:30:48.976486 18755 net.cpp:106] Creating Layer relu6 I0130 16:30:48.976495 18755 net.cpp:454] relu6 <- fc6 I0130 16:30:48.976505 18755 net.cpp:397] relu6 -> fc6 (in-place) I0130 16:30:48.977505 18755 net.cpp:150] Setting up relu6 I0130 16:30:48.977529 18755 net.cpp:157] Top shape: 1 4096 (4096) I0130 16:30:48.977533 18755 net.cpp:165] Memory required for data: 285226400 I0130 16:30:48.977540 18755 layer_factory.hpp:77] Creating layer drop6 I0130 16:30:48.977551 18755 net.cpp:106] Creating Layer drop6 I0130 16:30:48.977558 18755 net.cpp:454] drop6 <- fc6 I0130 16:30:48.977566 18755 net.cpp:397] drop6 -> fc6 (in-place) I0130 16:30:48.977615 18755 net.cpp:150] Setting up drop6 I0130 16:30:48.977622 18755 net.cpp:157] Top shape: 1 4096 (4096) I0130 16:30:48.977624 18755 net.cpp:165] Memory required for data: 285242784 I0130 16:30:48.977628 18755 layer_factory.hpp:77] Creating layer fc7 I0130 16:30:48.977635 18755 net.cpp:106] Creating Layer fc7 I0130 16:30:48.977638 18755 net.cpp:454] fc7 <- fc6 I0130 16:30:48.977643 18755 net.cpp:411] fc7 -> fc7 I0130 16:30:49.007350 18755 net.cpp:150] Setting up fc7 I0130 16:30:49.007377 18755 net.cpp:157] Top shape: 1 4096 (4096) I0130 16:30:49.007381 18755 net.cpp:165] Memory required for data: 285259168 I0130 16:30:49.007390 18755 layer_factory.hpp:77] Creating layer relu7 I0130 16:30:49.007400 18755 net.cpp:106] Creating Layer relu7 I0130 16:30:49.007403 18755 net.cpp:454] relu7 <- fc7 I0130 16:30:49.007410 18755 net.cpp:397] relu7 -> fc7 (in-place) I0130 16:30:49.007612 18755 net.cpp:150] Setting up relu7 I0130 16:30:49.007622 18755 net.cpp:157] Top shape: 1 4096 (4096) I0130 16:30:49.007623 18755 net.cpp:165] Memory required for data: 285275552 I0130 16:30:49.007627 18755 layer_factory.hpp:77] Creating layer drop7 I0130 16:30:49.007633 18755 net.cpp:106] Creating Layer drop7 I0130 16:30:49.007637 18755 net.cpp:454] drop7 <- fc7 I0130 16:30:49.007640 18755 net.cpp:397] drop7 -> fc7 (in-place) I0130 16:30:49.007668 18755 net.cpp:150] Setting up drop7 I0130 16:30:49.007673 18755 net.cpp:157] Top shape: 1 4096 (4096) I0130 16:30:49.007674 18755 net.cpp:165] Memory required for data: 285291936 I0130 16:30:49.007678 18755 layer_factory.hpp:77] Creating layer fc7_drop7_0_split I0130 16:30:49.007681 18755 net.cpp:106] Creating Layer fc7_drop7_0_split I0130 16:30:49.007684 18755 net.cpp:454] fc7_drop7_0_split <- fc7 I0130 16:30:49.007688 18755 net.cpp:411] fc7_drop7_0_split -> fc7_drop7_0_split_0 I0130 16:30:49.007692 18755 net.cpp:411] fc7_drop7_0_split -> fc7_drop7_0_split_1 I0130 16:30:49.007725 18755 net.cpp:150] Setting up fc7_drop7_0_split I0130 16:30:49.007730 18755 net.cpp:157] Top shape: 1 4096 (4096) I0130 16:30:49.007731 18755 net.cpp:157] Top shape: 1 4096 (4096) I0130 16:30:49.007733 18755 net.cpp:165] Memory required for data: 285324704 I0130 16:30:49.007736 18755 layer_factory.hpp:77] Creating layer cls_score I0130 16:30:49.007741 18755 net.cpp:106] Creating Layer cls_score I0130 16:30:49.007745 18755 net.cpp:454] cls_score <- fc7_drop7_0_split_0 I0130 16:30:49.007747 18755 net.cpp:411] cls_score -> cls_score I0130 16:30:49.007972 18755 net.cpp:150] Setting up cls_score I0130 16:30:49.007977 18755 net.cpp:157] Top shape: 1 2 (2) I0130 16:30:49.007979 18755 net.cpp:165] Memory required for data: 285324712 I0130 16:30:49.007984 18755 layer_factory.hpp:77] Creating layer bbox_pred I0130 16:30:49.007989 18755 net.cpp:106] Creating Layer bbox_pred I0130 16:30:49.007992 18755 net.cpp:454] bbox_pred <- fc7_drop7_0_split_1 I0130 16:30:49.007997 18755 net.cpp:411] bbox_pred -> bbox_pred I0130 16:30:49.009044 18755 net.cpp:150] Setting up bbox_pred I0130 16:30:49.009057 18755 net.cpp:157] Top shape: 1 10 (10) I0130 16:30:49.009058 18755 net.cpp:165] Memory required for data: 285324752 I0130 16:30:49.009063 18755 layer_factory.hpp:77] Creating layer loss_cls I0130 16:30:49.009070 18755 net.cpp:106] Creating Layer loss_cls I0130 16:30:49.009073 18755 net.cpp:454] loss_cls <- cls_score I0130 16:30:49.009078 18755 net.cpp:454] loss_cls <- labels I0130 16:30:49.009083 18755 net.cpp:411] loss_cls -> cls_loss I0130 16:30:49.009089 18755 layer_factory.hpp:77] Creating layer loss_cls I0130 16:30:49.009377 18755 net.cpp:150] Setting up loss_cls I0130 16:30:49.009385 18755 net.cpp:157] Top shape: (1) I0130 16:30:49.009388 18755 net.cpp:160] with loss weight 1 I0130 16:30:49.009399 18755 net.cpp:165] Memory required for data: 285324756 I0130 16:30:49.009402 18755 layer_factory.hpp:77] Creating layer loss_bbox I0130 16:30:49.009408 18755 net.cpp:106] Creating Layer loss_bbox I0130 16:30:49.009413 18755 net.cpp:454] loss_bbox <- bbox_pred I0130 16:30:49.009415 18755 net.cpp:454] loss_bbox <- bbox_targets I0130 16:30:49.009418 18755 net.cpp:454] loss_bbox <- bbox_inside_weights I0130 16:30:49.009421 18755 net.cpp:454] loss_bbox <- bbox_outside_weights I0130 16:30:49.009425 18755 net.cpp:411] loss_bbox -> bbox_loss I0130 16:30:49.009531 18755 net.cpp:150] Setting up loss_bbox I0130 16:30:49.009536 18755 net.cpp:157] Top shape: (1) I0130 16:30:49.009538 18755 net.cpp:160] with loss weight 1 I0130 16:30:49.009543 18755 net.cpp:165] Memory required for data: 285324760 I0130 16:30:49.009547 18755 net.cpp:226] loss_bbox needs backward computation. I0130 16:30:49.009551 18755 net.cpp:226] loss_cls needs backward computation. I0130 16:30:49.009553 18755 net.cpp:226] bbox_pred needs backward computation. I0130 16:30:49.009557 18755 net.cpp:226] cls_score needs backward computation. I0130 16:30:49.009559 18755 net.cpp:226] fc7_drop7_0_split needs backward computation. I0130 16:30:49.009562 18755 net.cpp:226] drop7 needs backward computation. I0130 16:30:49.009564 18755 net.cpp:226] relu7 needs backward computation. I0130 16:30:49.009567 18755 net.cpp:226] fc7 needs backward computation. I0130 16:30:49.009569 18755 net.cpp:226] drop6 needs backward computation. I0130 16:30:49.009572 18755 net.cpp:226] relu6 needs backward computation. I0130 16:30:49.009574 18755 net.cpp:226] fc6 needs backward computation. I0130 16:30:49.009577 18755 net.cpp:226] roi_pool_conv5 needs backward computation. I0130 16:30:49.009582 18755 net.cpp:226] roi-data needs backward computation. I0130 16:30:49.009585 18755 net.cpp:226] proposal needs backward computation. I0130 16:30:49.009589 18755 net.cpp:226] rpn_cls_prob_reshape needs backward computation. I0130 16:30:49.009593 18755 net.cpp:226] rpn_cls_prob needs backward computation. I0130 16:30:49.009595 18755 net.cpp:226] rpn_loss_bbox needs backward computation. I0130 16:30:49.009599 18755 net.cpp:226] rpn_loss_cls needs backward computation. I0130 16:30:49.009603 18755 net.cpp:226] rpn-data needs backward computation. I0130 16:30:49.009608 18755 net.cpp:226] rpn_cls_score_reshape_rpn_cls_score_reshape_0_split needs backward computation. I0130 16:30:49.009611 18755 net.cpp:226] rpn_cls_score_reshape needs backward computation. I0130 16:30:49.009614 18755 net.cpp:226] rpn_bbox_pred_rpn_bbox_pred_0_split needs backward computation. I0130 16:30:49.009618 18755 net.cpp:226] rpn_bbox_pred needs backward computation. I0130 16:30:49.009621 18755 net.cpp:226] rpn_cls_score_rpn_cls_score_0_split needs backward computation. I0130 16:30:49.009624 18755 net.cpp:226] rpn_cls_score needs backward computation. I0130 16:30:49.009627 18755 net.cpp:226] rpn/output_rpn_relu/3x3_0_split needs backward computation. I0130 16:30:49.009631 18755 net.cpp:226] rpn_relu/3x3 needs backward computation. I0130 16:30:49.009634 18755 net.cpp:226] rpn_conv/3x3 needs backward computation. I0130 16:30:49.009636 18755 net.cpp:226] conv5_relu5_0_split needs backward computation. I0130 16:30:49.009640 18755 net.cpp:226] relu5 needs backward computation. I0130 16:30:49.009644 18755 net.cpp:226] conv5 needs backward computation. I0130 16:30:49.009645 18755 net.cpp:226] relu4 needs backward computation. I0130 16:30:49.009649 18755 net.cpp:226] conv4 needs backward computation. I0130 16:30:49.009651 18755 net.cpp:226] relu3 needs backward computation. I0130 16:30:49.009654 18755 net.cpp:226] conv3 needs backward computation. I0130 16:30:49.009657 18755 net.cpp:226] pool2 needs backward computation. I0130 16:30:49.009660 18755 net.cpp:226] norm2 needs backward computation. I0130 16:30:49.009663 18755 net.cpp:226] relu2 needs backward computation. I0130 16:30:49.009665 18755 net.cpp:226] conv2 needs backward computation. I0130 16:30:49.009668 18755 net.cpp:226] pool1 needs backward computation. I0130 16:30:49.009671 18755 net.cpp:226] norm1 needs backward computation. I0130 16:30:49.009673 18755 net.cpp:226] relu1 needs backward computation. I0130 16:30:49.009676 18755 net.cpp:226] conv1 needs backward computation. I0130 16:30:49.009680 18755 net.cpp:228] gt_boxes_input-data_2_split does not need backward computation. I0130 16:30:49.009683 18755 net.cpp:228] im_info_input-data_1_split does not need backward computation. I0130 16:30:49.009686 18755 net.cpp:228] data_input-data_0_split does not need backward computation. I0130 16:30:49.009690 18755 net.cpp:228] input-data does not need backward computation. I0130 16:30:49.009692 18755 net.cpp:270] This network produces output bbox_loss I0130 16:30:49.009694 18755 net.cpp:270] This network produces output cls_loss I0130 16:30:49.009697 18755 net.cpp:270] This network produces output rpn_cls_loss I0130 16:30:49.009701 18755 net.cpp:270] This network produces output rpn_loss_bbox I0130 16:30:49.009732 18755 net.cpp:283] Network initialization done. I0130 16:30:49.009883 18755 solver.cpp:60] Solver scaffolding done. Loading pretrained model weights from /home/x306/wzy/py-faster-rcnn/data/imagenet_models/ZF.v2.caffemodel I0130 16:30:49.328835 18755 net.cpp:816] Ignoring source layer pool5_spm6 I0130 16:30:49.328855 18755 net.cpp:816] Ignoring source layer pool5_spm6_flatten I0130 16:30:49.385639 18755 net.cpp:816] Ignoring source layer fc8 I0130 16:30:49.385661 18755 net.cpp:816] Ignoring source layer prob I0130 16:30:50.523666 18755 solver.cpp:229] Iteration 0, loss = 1.41919 I0130 16:30:50.523718 18755 solver.cpp:245] Train net output #0: bbox_loss = 0.000346801 ( 1 = 0.000346801 loss) I0130 16:30:50.523727 18755 solver.cpp:245] Train net output #1: cls_loss = 0.539508 ( 1 = 0.539508 loss) I0130 16:30:50.523735 18755 solver.cpp:245] Train net output #2: rpn_cls_loss = 0.75761 ( 1 = 0.75761 loss) I0130 16:30:50.523741 18755 solver.cpp:245] Train net output #3: rpn_loss_bbox = 0 ( 1 = 0 loss) I0130 16:30:50.523749 18755 sgd_solver.cpp:106] Iteration 0, lr = 0.001 F0130 16:30:50.527855 18755 sgd_solver.cu:19] Check failed: error == cudaSuccess (11 vs. 0) invalid argument Check failure stack trace:

Process finished with exit code 134 (interrupted by signal 6: SIGABRT)

mjq11302010044 commented 6 years ago

@wzhiyuan2016 Sorry, but I notice that you are using ZF-net to train RRPN. But I think VGG16 can work well.