测试

[lhl2xju]

您好，我配置了环境，下载了您训练好的权重model_7709.pth，对比论文中的实验数据，但是我用这个权重文件测试iiit_str只得到了72.89的mAP，我不知道这是怎么回事，我不知道是环境还是其他的原因。

[lanfeng4659]

您好，我配置了环境，下载了您训练好的权重model_7709.pth，对比论文中的实验数据，但是我用这个权重文件测试iiit_str只得到了72.89的mAP，我不知道这是怎么回事，我不知道是环境还是其他的原因。

可以把测试的log.txt文件提供看一下吗

[lhl2xju]

因为我提问的时间较晚，以为您会隔日回复，实在抱歉，第一个是SVT数据集的log文件 第二个是IIIT-STR数据集的log文件（我将maskrcnn_benchmark包的名字改为了mnbk） [Uploading log_iiit10000.txt…]()

[lhl2xju]

log_svt249.txt

[lhl2xju]

log_iiit10000.txt

[lhl2xju]

log_iiit10000.txt

[lanfeng4659]

因为我提问的时间较晚，以为您会隔日回复，实在抱歉，第一个是SVT数据集的log文件 第二个是IIIT-STR数据集的log文件（我将maskrcnn_benchmark包的名字改为了mnbk） Uploading log_iiit10000.txt

以下是我测试的log，可以参考对比一下。

2023-06-27 17:55:55,377 maskrcnn_benchmark INFO: Using 1 GPUs 2023-06-27 17:55:55,377 maskrcnn_benchmark INFO: AMP_VERBOSE: False DARTS: ARCH_START_ITER: 5000 LR_A: 0.001 LR_END: 0.0001 TIE_CELL: False T_MAX: 2500 WD_A: 0.001 DATALOADER: ASPECT_RATIO_GROUPING: True NUM_WORKERS: 8 SIZE_DIVISIBILITY: 32 DATASETS: TEST: ('iiit_test',) TEXT: NUM_CHARS: 25 VOC_SIZE: 97 TRAIN: () DTYPE: float32 INPUT: AUGMENT: PSSAugmentation BRIGHTNESS: 0.125 CONTRAST: 0.125 CROP_PROB_TRAIN: 0.0 CROP_SIZE_TRAIN: -1 FLIP_PROB_TRAIN: 0.0 HUE: 0.5 MAX_SIZE_TEST: 1333 MAX_SIZE_TRAIN: 1333 MIN_SIZE_RANGE_TRAIN: (-1, -1) MIN_SIZE_TEST: 800 MIN_SIZE_TRAIN: (800,) PIXEL_MEAN: [103.53, 116.28, 123.675] PIXEL_STD: [57.375, 57.12, 58.395] SATURATION: 0.5 TO_BGR255: True VERTICAL_FLIP_PROB_TRAIN: 0.0 IS_LOAD_OPTIMIZER: True IS_LOAD_SCHEDULER: True MODEL: ALIGN: IS_CHINESE: False NUM_CONVS: 2 POOLER_CANONICAL_SCALE: 160 POOLER_RESOLUTION: (4, 15) POOLER_SCALES: (0.25, 0.125, 0.0625) PREDICTOR: ctc PYRAMID_LAYERS: (2, 3, 4, 5) USE_ALONG_LOSS: False USE_BOX_AUG: False USE_CHARACTER_AWARENESS: False USE_CHAR_COUNT: False USE_COMMON_SPACE: False USE_CONTRASTIVE_LOSS: False USE_CTC_LOSS: False USE_DOMAIN_ALIGN_LOSS: False USE_DOMAIN_CLASSIFIER: False USE_DYNAMIC_SIMILARITY: False USE_FOCAL_L1_LOSS: False USE_GLOBAL_LOCAL_SIMILARITY: False USE_HANMING: False USE_IOU_PREDICTOR: False USE_LOOK_UP: False USE_NO_RNN: False USE_N_GRAM_ED: False USE_PYRAMID: False USE_RES_LINK: False USE_RETRIEVAL: True USE_STEP: False USE_TEXTNESS: False USE_WORD_AUG: False USE_WORD_INSTANCE_AUG: False ATTENTION: IS_CHINESE: False NUM_CONVS: 4 POOLER_CANONICAL_SCALE: 160 POOLER_RESOLUTION: (14, 64) POOLER_SCALES: (0.125, 0.0625, 0.03125) PREDICTOR: ctc USE_BOX_AUG: False USE_RETRIEVAL: True USE_WORD_AUG: False BACKBONE: CONV_BODY: R-50 FREEZE_BN: False FREEZE_CONV_BODY_AT: 2 CHAR_INST_ON: False CHAR_ON: False CLS_AGNOSTIC_BBOX_REG: False DEVICE: cuda EAST: CENTER_SAMPLE: True FPN_STRIDES: [8, 16, 32, 64, 128] INFERENCE_TH: 0.2 LOC_LOSS_TYPE: giou LOSS_ALPHA: 0.25 LOSS_GAMMA: 2.0 NMS_TH: 0.6 NUM_CLASSES: 2 NUM_CONVS: 2 POS_RADIUS: 1.5 PRE_NMS_TOP_N: 1000 PRIOR_PROB: 0.01 SIZES_OF_INTEREST: [64, 128, 256, 512] USE_BN: False USE_DEFORMABLE: False USE_GN: True USE_LIGHTWEIGHT: False USE_RELU: True FBNET: ARCH: default ARCH_DEF: BN_TYPE: bn DET_HEAD_BLOCKS: [] DET_HEAD_LAST_SCALE: 1.0 DET_HEAD_STRIDE: 0 DW_CONV_SKIP_BN: True DW_CONV_SKIP_RELU: True KPTS_HEAD_BLOCKS: [] KPTS_HEAD_LAST_SCALE: 0.0 KPTS_HEAD_STRIDE: 0 MASK_HEAD_BLOCKS: [] MASK_HEAD_LAST_SCALE: 0.0 MASK_HEAD_STRIDE: 0 RPN_BN_TYPE: RPN_HEAD_BLOCKS: 0 SCALE_FACTOR: 1.0 WIDTH_DIVISOR: 1 FCOS: CENTER_SAMPLE: True FPN_STRIDES: [8, 16, 32, 64, 128] INFERENCE_TH: 0.05 LOC_LOSS_TYPE: giou LOSS_ALPHA: 0.25 LOSS_GAMMA: 2.0 NMS_TH: 0.6 NUM_CLASSES: 2 NUM_CONVS: 4 POS_RADIUS: 1.5 PRE_NMS_TOP_N: 1000 PRIOR_PROB: 0.01 SIZES_OF_INTEREST: [64, 128, 256, 512] USE_BN: False USE_DEFORMABLE: False USE_GN: True USE_LIGHTWEIGHT: False USE_RELU: True FCOS_ON: True FPN: USE_BN: False USE_DEFORMABLE: False USE_GN: False USE_RELU: False GROUP_NORM: DIM_PER_GP: -1 EPSILON: 1e-05 NUM_GROUPS: 32 HEAD: BBOX_LOSS: ALPHA: 0.5 BETA: 0.11 GAMMA: 1.5 TYPE: IOULoss WEIGHT: 1.0 INST_ON: False KEYPOINT_ON: False KE_ON: False MASK_ON: False META_ARCHITECTURE: OneStage MSR_ON: False NECK: CONV_BODY: fpn-align IN_CHANNELS: 256 LAST_STRIDE: 2 NUM_LEVELS: 5 REFINE_LEVEL: 1 REFINE_TYPE: non_local USE_DEFORMABLE: False USE_GN: False OFFSET: KERNEL_SIZE: 3 PREDICTOR: polar STOP_OFFSETS: 1500 ONE_STAGE_HEAD: align POLYGON_DET: False RESNETS: BACKBONE_OUT_CHANNELS: 256 DEFORMABLE_GROUPS: 1 DEFORM_POOLING: False MAX_DCN_LAYER: 15 NUM_GROUPS: 1 RES2_OUT_CHANNELS: 256 RES5_DILATION: 1 STAGE_WITH_CONTEXT: (False, False, False, False) STAGE_WITH_DCN: (False, False, False, False) STEM_FUNC: StemWithFixedBatchNorm STEM_OUT_CHANNELS: 64 STRIDE_IN_1X1: True TRANS_FUNC: BottleneckWithFixedBatchNorm WIDTH_PER_GROUP: 64 WITH_MODULATED_DCN: False RETINANET: ANCHOR_SIZES: (32, 64, 128, 256, 512) ANCHOR_STRIDES: (8, 16, 32, 64, 128) ASPECT_RATIOS: (0.5, 1.0, 2.0) BBOX_REG_BETA: 0.11 BBOX_REG_WEIGHT: 4.0 BG_IOU_THRESHOLD: 0.4 FG_IOU_THRESHOLD: 0.5 INFERENCE_TH: 0.05 LOSS_ALPHA: 0.25 LOSS_GAMMA: 2.0 NMS_TH: 0.4 NUM_CLASSES: 81 NUM_CONVS: 4 OCTAVE: 2.0 PRE_NMS_TOP_N: 1000 PRIOR_PROB: 0.01 SCALES_PER_OCTAVE: 3 STRADDLE_THRESH: 0 USE_C5: False RETINANET_ON: False RETRIEVAL_ONLY: False ROI_BOX_HEAD: CLASS_WEIGHT: 1.0 CONV_HEAD_DIM: 256 DILATION: 1 FEATURE_EXTRACTOR: ResNet50Conv5ROIFeatureExtractor MLP_HEAD_DIM: 1024 NUM_CLASSES: 81 NUM_STACKED_CONVS: 4 POOLER_RESOLUTION: 14 POOLER_SAMPLING_RATIO: 0 POOLER_SCALES: (0.0625,) PREDICTOR: FastRCNNPredictor USE_DFPOOL: False USE_GN: False ROI_HEADS: BATCH_SIZE_PER_IMAGE: 512 BBOX_REG_WEIGHTS: (10.0, 10.0, 5.0, 5.0) BG_IOU_THRESHOLD: 0.5 DETECTIONS_PER_IMG: 100 FG_IOU_THRESHOLD: 0.5 NMS: 0.5 POSITIVE_FRACTION: 0.25 SCORE_THRESH: 0.05 USE_FPN: False ROI_INST_HEAD: PREDICTOR: EmbeddingPredictor ROI_KEYPOINT_HEAD: CONV_LAYERS: (512, 512, 512, 512, 512, 512, 512, 512) FEATURE_EXTRACTOR: KeypointRCNNFeatureExtractor MLP_HEAD_DIM: 1024 NUM_CLASSES: 17 POOLER_RESOLUTION: 14 POOLER_SAMPLING_RATIO: 0 POOLER_SCALES: (0.0625,) PREDICTOR: KeypointRCNNPredictor RESOLUTION: 14 SHARE_BOX_FEATURE_EXTRACTOR: True ROI_MASK_HEAD: CONV_LAYERS: (256, 256, 256, 256) DILATION: 1 FEATURE_EXTRACTOR: ResNet50Conv5ROIFeatureExtractor MLP_HEAD_DIM: 1024 POOLER_RESOLUTION: 14 POOLER_SAMPLING_RATIO: 0 POOLER_SCALES: (0.0625,) POSTPROCESS_MASKS: False POSTPROCESS_MASKS_THRESHOLD: 0.5 PREDICTOR: MaskRCNNC4Predictor RESOLUTION: 14 SHARE_BOX_FEATURE_EXTRACTOR: True USE_DFPOOL: False USE_GN: False RPN: ANCHOR_SIZES: (32, 64, 128, 256, 512) ANCHOR_STRIDE: (16,) ASPECT_RATIOS: (0.5, 1.0, 2.0) BATCH_SIZE_PER_IMAGE: 256 BG_IOU_THRESHOLD: 0.3 FG_IOU_THRESHOLD: 0.7 FPN_POST_NMS_PER_BATCH: True FPN_POST_NMS_TOP_N_TEST: 2000 FPN_POST_NMS_TOP_N_TRAIN: 2000 MIN_SIZE: 0 NMS_THRESH: 0.7 POSITIVE_FRACTION: 0.5 POST_NMS_TOP_N_TEST: 1000 POST_NMS_TOP_N_TRAIN: 2000 PRE_NMS_TOP_N_TEST: 6000 PRE_NMS_TOP_N_TRAIN: 12000 RPN_HEAD: SingleConvRPNHead STRADDLE_THRESH: 0 USE_FPN: False RPN_ONLY: True WEIGHT: catalog://ImageNetPretrained/MSRA/R-50 OUTPUT_DIR: Log/evluation PATHS_CATALOG: /workspace/wanghao/projects/STR-TDSL/maskrcnn_benchmark/config/paths_catalog.py PROCESS: NMS_THRESH: 0.4 PNMS: False SOLVER: BASE_LR: 0.001 BIAS_LR_FACTOR: 2 CHECKPOINT_PERIOD: 2500 GAMMA: 0.1 IMS_PER_BATCH: 16 MAX_ITER: 40000 MOMENTUM: 0.9 ONE_STAGE_HEAD_LR_FACTOR: 1.0 POLY_POWER: 0.9 SCHEDULER: multistep STEPS: (30000,) WARMUP_FACTOR: 0.3333333333333333 WARMUP_ITERS: 500 WARMUP_METHOD: linear WEIGHT_DECAY: 0.0005 WEIGHT_DECAY_BIAS: 0 SYNCBN: False TEST: BBOX_AUG: ENABLED: False H_FLIP: False MAX_SIZE: 4000 SCALES: () SCALE_H_FLIP: False DETECTIONS_PER_IMG: 100 EXPECTED_RESULTS: [] EXPECTED_RESULTS_SIGMA_TOL: 4 IMS_PER_BATCH: 1 2023-06-27 17:55:55,377 maskrcnn_benchmark INFO: Collecting env info (might take some time) 2023-06-27 17:56:00,256 maskrcnn_benchmark INFO: PyTorch version: 1.2.0 Is debug build: No CUDA used to build PyTorch: 10.0.130

OS: Ubuntu 20.04.4 LTS GCC version: (Ubuntu 7.5.0-6ubuntu2) 7.5.0 CMake version: Could not collect

Python version: 3.7 Is CUDA available: Yes CUDA runtime version: 10.2.89 GPU models and configuration: GPU 0: NVIDIA TITAN X (Pascal) GPU 1: NVIDIA TITAN X (Pascal) GPU 2: NVIDIA TITAN X (Pascal) GPU 3: NVIDIA TITAN X (Pascal)

Nvidia driver version: 470.182.03 cuDNN version: /usr/lib/x86_64-linux-gnu/libcudnn.so.7.6.5

Versions of relevant libraries: [pip3] numpy==1.18.5 [pip3] torch==1.2.0 [pip3] torchaudio==0.8.0 [pip3] torchvision==0.4.0 [conda] blas 1.0 mkl https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main [conda] mkl 2020.1 217 https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main [conda] mkl-service 2.3.0 py37he904b0f_0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main [conda] mkl_fft 1.1.0 py37h23d657b_0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main [conda] mkl_random 1.1.1 py37h0573a6f_0 https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main [conda] torch 1.2.0 pypi_0 pypi [conda] torchaudio 0.8.0 pypi_0 pypi [conda] torchvision 0.4.0 pypi_0 pypi Pillow (7.2.0) <class 'maskrcnn_benchmark.data.datasets.iiit.IIITDataset'> 2023-06-27 17:56:05,592 maskrcnn_benchmark.utils.checkpoint INFO: Loading checkpoint from model_rec_synth_ic17_7709.pth 2023-06-27 17:56:06,320 maskrcnn_benchmark.inference INFO: Start evaluation on iiit_test dataset(10000 images). 2023-06-27 18:15:12,922 maskrcnn_benchmark.inference INFO: Total run time: 0:19:06.601513 (0.1146601513147354 s / img per device, on 1 devices) 2023-06-27 18:15:12,922 maskrcnn_benchmark.inference INFO: Model inference time: 0:15:10.037243 (0.0910037243127823 s / img per device, on 1 devices) 2023-06-27 18:15:45,493 maskrcnn_benchmark.inference INFO: Evaluating bbox proposals Model:model_rec_synth_ic17_7709.pth,mAP:0.7709092824651216,best mAP:0.7709092824651216

[lanfeng4659]

您好，我配置了环境，下载了您训练好的权重model_7709.pth，对比论文中的实验数据，但是我用这个权重文件测试iiit_str只得到了72.89的mAP，我不知道这是怎么回事，我不知道是环境还是其他的原因。

您好，问题解决了告知一声。如果是代码问题，我好更新。

[lhl2xju]

好的，非常感谢您，我去配置一下您的测试环境，然后再测试一遍，测试结果出来再留言告知您

[lhl2xju]

我对比了您的配置文件重新测试了一遍，结果和上一次的一样，您可以更新一下代码然后我再去测试一下，因为这可能是我配置的虚拟环境出了问题

[lhl2xju]

您好，我用了一些取巧的方法重新测试了一遍，实验结果和您论文中描述的一样，非常感谢您的解答