search

PaddlePaddle / PaddleDetection

Object Detection toolkit based on PaddlePaddle. It supports object detection, instance segmentation, multiple object tracking and real-time multi-person keypoint detection.
Apache License 2.0
12.84k stars 2.89k forks source link

faster_rcnn_r50_vd_fpn_ciou_loss_1x运行VOC数据集上报错 #787

Closed yinggo closed 4 years ago

yinggo commented 4 years ago 
2020-05-26 23:48:12,707-INFO: Save model to output/faster_rcnn_r50_vd_fpn_ciou_loss_1x/10000.
I0526 23:48:15.279527 24507 parallel_executor.cc:440] The Program will be executed on CUDA using ParallelExecutor, 1 cards are used, so 1 programs are executed in parallel.
I0526 23:48:15.287273 24507 build_strategy.cc:365] SeqOnlyAllReduceOps:0, num_trainers:1
W0526 23:48:15.904718 24683 operator.cc:181] py_func raises an exception pybind11::error_already_set, ValueError: all the input arrays must have same number of dimensions, but the array at index 0 has 2 dimension(s) and the array at index 1 has 3 dimension(s)

At:
  <__array_function__ internals>(6): concatenate
  /home/ds1/anaconda3/envs/paddle/lib/python3.7/site-packages/numpy/core/shape_base.py(344): hstack
  <__array_function__ internals>(6): hstack
  /home/ds1/anaconda3/envs/paddle/PaddleDetection/ppdet/modeling/ops.py(718): _diou_nms
  /home/ds1/anaconda3/envs/paddle/lib/python3.7/site-packages/paddle/fluid/layers/nn.py(12409): __call__
F0526 23:48:15.904808 24683 exception_holder.h:37] std::exception caught, ValueError: all the input arrays must have same number of dimensions, but the array at index 0 has 2 dimension(s) and the array at index 1 has 3 dimension(s)

At:
  <__array_function__ internals>(6): concatenate
  /home/ds1/anaconda3/envs/paddle/lib/python3.7/site-packages/numpy/core/shape_base.py(344): hstack
  <__array_function__ internals>(6): hstack
  /home/ds1/anaconda3/envs/paddle/PaddleDetection/ppdet/modeling/ops.py(718): _diou_nms
  /home/ds1/anaconda3/envs/paddle/lib/python3.7/site-packages/paddle/fluid/layers/nn.py(12409): __call__
*** Check failure stack trace: ***
    @     0x7f81fe916c2d  google::LogMessage::Fail()
    @     0x7f81fe91a6dc  google::LogMessage::SendToLog()
    @     0x7f81fe916753  google::LogMessage::Flush()
    @     0x7f81fe91bbee  google::LogMessageFatal::~LogMessageFatal()
    @     0x7f8200ef09b8  paddle::framework::details::ExceptionHolder::Catch()
    @     0x7f8200f9c68e  paddle::framework::details::FastThreadedSSAGraphExecutor::RunOpSync()
    @     0x7f8200f9b29f  paddle::framework::details::FastThreadedSSAGraphExecutor::RunOp()
    @     0x7f8200f9b564  _ZNSt17_Function_handlerIFvvESt17reference_wrapperISt12_Bind_simpleIFS1_ISt5_BindIFZN6paddle9framework7details28FastThreadedSSAGraphExecutor10RunOpAsyncEPSt13unordered_mapIPNS6_12OpHandleBaseESt6atomicIiESt4hashISA_ESt8equal_toISA_ESaISt4pairIKSA_SC_EEESA_RKSt10shared_ptrINS5_13BlockingQueueImEEEEUlvE_vEEEvEEEE9_M_invokeERKSt9_Any_data
    @     0x7f81fe96f983  std::_Function_handler<>::_M_invoke()
    @     0x7f81fe6fdc37  std::__future_base::_State_base::_M_do_set()
    @     0x7f8231911a99  __pthread_once_slow
    @     0x7f8200f96a52  _ZNSt13__future_base11_Task_stateISt5_BindIFZN6paddle9framework7details28FastThreadedSSAGraphExecutor10RunOpAsyncEPSt13unordered_mapIPNS4_12OpHandleBaseESt6atomicIiESt4hashIS8_ESt8equal_toIS8_ESaISt4pairIKS8_SA_EEES8_RKSt10shared_ptrINS3_13BlockingQueueImEEEEUlvE_vEESaIiEFvvEE6_M_runEv
    @     0x7f81fe6ffe64  _ZZN10ThreadPoolC1EmENKUlvE_clEv
    @     0x7f822a83a3e7  execute_native_thread_routine_compat
    @     0x7f823190a6ba  start_thread
    @     0x7f823164041d  clone
    @              (nil)  (unknown)

之前也测试过一些模型，都是在VOC数据格式下做的，但这个就跑不通。。求指教

（贴一下config）
：
architecture: FasterRCNN
max_iters: 90000
snapshot_iter: 10000
use_gpu: true
log_smooth_window: 20
save_dir: output
pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_vd_pretrained.tar
weights: output/faster_rcnn_r50_vd_fpn_ciou_loss_1x_voc/model_final
metric: VOC
num_classes: 20

FasterRCNN:
  backbone: ResNet
  fpn: FPN
  rpn_head: FPNRPNHead
  roi_extractor: FPNRoIAlign
  bbox_head: BBoxHead
  bbox_assigner: BBoxAssigner

ResNet:
  depth: 50
  feature_maps: [2, 3, 4, 5]
  freeze_at: 2
  norm_type: bn
  variant: d

FPN:
  max_level: 6
  min_level: 2
  num_chan: 256
  spatial_scale: [0.03125, 0.0625, 0.125, 0.25]

FPNRPNHead:
  anchor_generator:
    anchor_sizes: [32, 64, 128, 256, 512]
    aspect_ratios: [0.5, 1.0, 2.0]
    stride: [16.0, 16.0]
    variance: [1.0, 1.0, 1.0, 1.0]
  anchor_start_size: 32
  max_level: 6
  min_level: 2
  num_chan: 256
  rpn_target_assign:
    rpn_batch_size_per_im: 256
    rpn_fg_fraction: 0.5
    rpn_negative_overlap: 0.3
    rpn_positive_overlap: 0.7
    rpn_straddle_thresh: 0.0
  train_proposal:
    min_size: 0.0
    nms_thresh: 0.7
    post_nms_top_n: 2000
    pre_nms_top_n: 2000
  test_proposal:
    min_size: 0.0
    nms_thresh: 0.7
    post_nms_top_n: 1000
    pre_nms_top_n: 1000

FPNRoIAlign:
  canconical_level: 4
  canonical_size: 224
  max_level: 5
  min_level: 2
  box_resolution: 7
  sampling_ratio: 2

BBoxAssigner:
  batch_size_per_im: 512
  bbox_reg_weights: [0.1, 0.1, 0.2, 0.2]
  bg_thresh_hi: 0.5
  bg_thresh_lo: 0.0
  fg_fraction: 0.25
  fg_thresh: 0.5

BBoxHead:
  head: TwoFCHead
  nms: MultiClassDiouNMS
  bbox_loss: DiouLoss

MultiClassDiouNMS:
  keep_top_k: 100
  nms_threshold: 0.5
  score_threshold: 0.05

DiouLoss:
  loss_weight: 10.0
  is_cls_agnostic: false
  use_complete_iou_loss: true

TwoFCHead:
  mlp_dim: 1024

LearningRate:
  base_lr: 0.02
  schedulers:
  - !PiecewiseDecay
    gamma: 0.1
    milestones: [60000, 80000]
  - !LinearWarmup
    start_factor: 0.1
    steps: 1000

OptimizerBuilder:
  optimizer:
    momentum: 0.9
    type: Momentum
  regularizer:
    factor: 0.0001
    type: L2

_READER_: '../faster_fpn_reader.yml'
TrainReader:
  batch_size: 2

TrainReader:
  inputs_def:
    fields: ['image', 'im_info', 'im_id', 'gt_bbox', 'gt_class', 'is_crowd']
  dataset:
    !VOCDataSet
    anno_path: train.txt #annotations/instances_train2017.json
    dataset_dir: dataset/voc
    use_default_label: true
    with_background: false
  sample_transforms:
  - !DecodeImage
    to_rgb: true
  - !RandomFlipImage
    prob: 0.5
  - !NormalizeImage
    is_channel_first: false
    is_scale: True
    mean:
    - 0.485
    - 0.456
    - 0.406
    std:
    - 0.229
    - 0.224
    - 0.225
  - !ResizeImage
    interp: 1
    target_size: [416, 448, 480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800, 832, 864, 896, 928, 960, 992, 1024, 1056, 1088, 1120, 1152, 1184, 1216, 1248, 1280, 1312, 1344, 1376, 1408]
    max_size: 1800
    use_cv2: true
  - !Permute
    to_bgr: false
  batch_transforms:
  - !PadBatch
    pad_to_stride: 32
  batch_size: 2
  shuffle: true
  drop_last: false
  worker_num: 2

EvalReader:
  inputs_def:
    fields: ['image', 'im_info', 'im_id', 'im_shape', 'gt_bbox', 'gt_class', 'is_difficult']
    use_flip: true
  dataset:
    !VOCDataSet
    anno_path: val.txt #annotations/instances_val2017.json
    dataset_dir: dataset/voc
    use_default_label: true
    with_background: false
  sample_transforms:
  - !DecodeImage
    to_rgb: True
    with_mixup: False
  - !NormalizeImage
    is_channel_first: false
    is_scale: True
    mean:
    - 0.485
    - 0.456
    - 0.406
    std:
    - 0.229
    - 0.224
    - 0.225
  - !ResizeImage
    interp: 1
    target_size:
    - 1200
    max_size: 2000
    use_cv2: true
  - !Permute
    to_bgr: false
  batch_transforms:
  - !PadBatch
    pad_to_stride: 32
  batch_size: 2
  worker_num: 2
  drop_empty: false

TestReader:
  inputs_def:
    fields: ['image', 'im_info', 'im_id','im_shape']
  dataset:
    !ImageFolder
#    anno_path: annotations/instances_val2017.json
    use_default_label: true
    with_background: false
  sample_transforms:
  - !DecodeImage
    to_rgb: true
    with_mixup: false
  - !NormalizeImage
    is_channel_first: false
    is_scale: true
    mean: [0.485,0.456,0.406]
    std: [0.229, 0.224,0.225]
  - !ResizeImage
    interp: 1
    max_size: 1333
    target_size: 800
    use_cv2: true
  - !Permute
    channel_first: true
    to_bgr: false
  batch_transforms:
  - !PadBatch
    pad_to_stride: 32
    use_padded_im_info: true
  batch_size: 1
  worker_num: 2
jerrywgz commented 4 years ago

我在coco数据集上验证了下faster_rcnn_r50_vd_fpn_ciou_loss_1x是能够正常跑的。由于是rcnn系列模型，你的配置文件中的num_class需要增加背景类，同时在dataset中将with_background设置为true

yinggo commented 4 years ago

@jerrywgz 您好，我增加了背景类测试了一下 num_classes: 20 TrainReader:和EvalReader下修改 with_background: true 但还是出现同样的报错额…… 是否我上述配置文件中有其他地方修改不当？

jerrywgz commented 4 years ago

我用这个模型在voc数据集上做了下适配，训练预测是可以跑通的，使用的release/0.3版本的检测库

architecture: FasterRCNN
max_iters: 90000
snapshot_iter: 10000
use_gpu: true
log_smooth_window: 20
save_dir: output
pretrain_weights: https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_vd_pretrained.tar
weights: output/faster_rcnn_r50_vd_fpn_diou_loss_1x/model_final
metric: VOC 
num_classes: 21

FasterRCNN:
  backbone: ResNet
  fpn: FPN
  rpn_head: FPNRPNHead
  roi_extractor: FPNRoIAlign
  bbox_head: BBoxHead
  bbox_assigner: BBoxAssigner

ResNet:
  depth: 50
  feature_maps: [2, 3, 4, 5]
  freeze_at: 2
  norm_type: bn
  variant: d

FPN:
  max_level: 6
  min_level: 2
  num_chan: 256
  spatial_scale: [0.03125, 0.0625, 0.125, 0.25]

FPNRPNHead:
  anchor_generator:
    anchor_sizes: [32, 64, 128, 256, 512]
    aspect_ratios: [0.5, 1.0, 2.0]
    stride: [16.0, 16.0]
    variance: [1.0, 1.0, 1.0, 1.0]
  anchor_start_size: 32
  max_level: 6
  min_level: 2
  num_chan: 256
  rpn_target_assign:
    rpn_batch_size_per_im: 256
    rpn_fg_fraction: 0.5
    rpn_negative_overlap: 0.3
    rpn_positive_overlap: 0.7
    rpn_straddle_thresh: 0.0
  train_proposal:
    min_size: 0.0
    nms_thresh: 0.7
    post_nms_top_n: 2000
    pre_nms_top_n: 2000
  test_proposal:
    min_size: 0.0
    nms_thresh: 0.7
    post_nms_top_n: 1000
    pre_nms_top_n: 1000

FPNRoIAlign:
  canconical_level: 4
  canonical_size: 224
  max_level: 5
  min_level: 2
  box_resolution: 7
  sampling_ratio: 2

BBoxAssigner:
  batch_size_per_im: 512
  bbox_reg_weights: [0.1, 0.1, 0.2, 0.2]
  bg_thresh_hi: 0.5
  bg_thresh_lo: 0.0
  fg_fraction: 0.25
  fg_thresh: 0.5

BBoxHead:
  head: TwoFCHead
  nms: MultiClassDiouNMS
  bbox_loss: DiouLoss

MultiClassDiouNMS:
  keep_top_k: 100
  nms_threshold: 0.5
  score_threshold: 0.05

DiouLoss:
  loss_weight: 10.0
  is_cls_agnostic: false
  use_complete_iou_loss: true

TwoFCHead:
  mlp_dim: 1024

LearningRate:
  base_lr: 0.02
  schedulers:
  - !PiecewiseDecay
    gamma: 0.1
    milestones: [60000, 80000]
  - !LinearWarmup
    start_factor: 0.1
    steps: 1000

OptimizerBuilder:
  optimizer:
    momentum: 0.9
    type: Momentum
  regularizer:
    factor: 0.0001
    type: L2

_READER_: '../faster_fpn_reader.yml'
TrainReader:
  batch_size: 2
  dataset:
    !VOCDataSet
      dataset_dir: /paddle/wangguanzhong/data/voc
      anno_path: trainval.txt
      with_background: true
      use_default_label: true

EvalReader:
  inputs_def:
    fields: ['image', 'im_info', 'im_id', 'im_shape', 'gt_bbox' , 'gt_class', 'is_difficult']
  dataset:
    !VOCDataSet
    dataset_dir: /paddle/wangguanzhong/data/voc
    anno_path: test.txt
    with_background: true
    use_default_label: true
yinggo commented 4 years ago

感谢！已解决