PaddleDetection

PaddleDetection copied to clipboard

问题确认 Search before asking

• [X] 我已经查询历史issue，没有报过同样bug。I have searched the issues and found no similar bug report.

bug描述 Describe the Bug

PPhuman 使用 --run_mode trt_fp16 参数，程序 运行到Model Configuration ，不再往下运行，也不报错，也不返回

如果不使用 TensorRT，即不加--run_mode trt_fp16 参数，则可正常运行

复现环境 Environment

-paddlepaddle-gpu 2.2.2.post101 -paddledet 2.4.0 -python 3.8.0 -CUDA 10.1 -CUDNN 7.6.5 -OS WIN10

python deploy/pphuman/pipeline.py --config deploy/pphuman/config/infer_cfg.yml --video_file=G:\win_project\FairMOT\code\data_prepaired\2.mp4 --device=gpu --enable_action=True --run_mode trt_fp16 F:\python_project\paddle_detection_project\venv\lib\site-packages\paddle\vision\transforms\functional_pil.py:36: DeprecationWarning: N EAREST is deprecated and will be removed in Pillow 10 (2023-07-01). Use Resampling.NEAREST or Dither.NONE instead. 'nearest': Image.NEAREST, F:\python_project\paddle_detection_project\venv\lib\site-packages\paddle\vision\transforms\functional_pil.py:37: DeprecationWarning: B ILINEAR is deprecated and will be removed in Pillow 10 (2023-07-01). Use Resampling.BILINEAR instead. 'bilinear': Image.BILINEAR, F:\python_project\paddle_detection_project\venv\lib\site-packages\paddle\vision\transforms\functional_pil.py:38: DeprecationWarning: B ICUBIC is deprecated and will be removed in Pillow 10 (2023-07-01). Use Resampling.BICUBIC instead. 'bicubic': Image.BICUBIC, F:\python_project\paddle_detection_project\venv\lib\site-packages\paddle\vision\transforms\functional_pil.py:39: DeprecationWarning: B OX is deprecated and will be removed in Pillow 10 (2023-07-01). Use Resampling.BOX instead. 'box': Image.BOX, F:\python_project\paddle_detection_project\venv\lib\site-packages\paddle\vision\transforms\functional_pil.py:40: DeprecationWarning: L ANCZOS is deprecated and will be removed in Pillow 10 (2023-07-01). Use Resampling.LANCZOS instead. 'lanczos': Image.LANCZOS, F:\python_project\paddle_detection_project\venv\lib\site-packages\paddle\vision\transforms\functional_pil.py:41: DeprecationWarning: H AMMING is deprecated and will be removed in Pillow 10 (2023-07-01). Use Resampling.HAMMING instead. 'hamming': Image.HAMMING deploy/pphuman/pipeline.py:26: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated since Python 3.3, and in 3.9 it will stop working from collections import Sequence ----------- Running Arguments ----------- ACTION: batch_size: 1 coord_size:

• 384
• 512 display_frames: 80 max_frames: 50 model_dir: output_inference/STGCN ATTR: batch_size: 8 model_dir: output_inference/strongbaseline_r50_30e_pa100k/ DET: batch_size: 1 model_dir: output_inference/mot_ppyoloe_l_36e_pipeline/ KPT: batch_size: 8 model_dir: output_inference/dark_hrnet_w32_256x192/ MOT: batch_size: 1 model_dir: output_inference/mot_ppyoloe_l_36e_pipeline/ tracker_config: deploy/pphuman/config/tracker_config.yml REID: batch_size: 16 model_dir: output_inference/reid_model/ attr_thresh: 0.5 crop_thresh: 0.5 kpt_thresh: 0.2 visual: true warmup_frame: 50

---

Action Recognition enabled ----------- Model Configuration ----------- Model Arch: YOLO Transform Order: --transform op: Resize --transform op: Permute

程序停止在此处，无报错，无返回

是否愿意提交PR Are you willing to submit a PR?

• [ ] Yes I'd like to help by submitting a PR!

大约1个小时以后，模型加载完成。（太慢了） 但是使用TensorRT并没有加速，性能和不使用TensorRT相同

你好，非常感谢你的异常反馈，初步怀疑是对tensorrt的支持有些问题，我们先排查一下

你好，非常感谢你的异常反馈，初步怀疑是对tensorrt的支持有些问题，我们先排查一下

你好，想问一下tensorrt转换后的模型会存放在哪里，重新运行的时候又加载tensorrt的转换。 我的CUDA=10.2，cudnn=7.6.5，paddle是带trt的对应的版本，trt版本是7.0也是对应的

我也是你这样的问题，没有任何返回就自动结束了，

看看这个TensorRT 部署效率太差的BUG！。 另外，可以看看官网使用Paddle-TensorRT库预测资料。 我个人RTX3080上的测试结果是run_mode=trt_fp16会比run_mode=paddle快约2.5倍。

------------------ Inference Time Info ----------------------
total_time(ms): 10305.0, img_num: 582
mot time(ms): 8617.4; per frame average time(ms): 14.806529209621992
cls_action time(ms): 1637.3; per trackid average time(ms): 2.813230240549828
average latency time(ms): 17.71, QPS: 56.477438