kbyran/poinet

Person of Interest. A flexible computer vision library on human analysis, such as person re-identification, human attribute, pose estimation and dense pose estimation.

POI

Person of Interest. A flexible computer vision library on human analysis, such as person re-identification, human attribute, pose estimation and dense pose estimation.

Features

• All in One. Multiple tasks on human analysis are integrated into the one framework, relieving the trouble of maintaining several repositories.
• Easy to Develop. Simple baselines for each task are already implemented, and convenient for secondary development and research.
• Fast to Deploy. Support export deployment model on ONNX/TensorRT to make more devices compatible and particularly gain 50% speed-up on NVIDIA GPU.

Preparation

• INSTALL.md for installation.
• DATASET.md for dataset.
• PRETRAINED.md for pretrained model.

Getting Started

Please read GETTING_STARTED.md for hands-on tutorial.

Model Zoo

Please check MODEL_ZOO.md for models.

Visualization

Please refer to VISUALIZE.md.

License & disclaimer

This project is under the Apache 2.0 license. The framework is based on simpledet. Thanks for their great work!

poinet's Issues

Dataset query- PETA dataset instead of RAP dataset for person attribute recognition

Thanks for code. For Person attributes recognition can we use PETA dataset instead of RAP dataset. If so where should the change be made

Inference failed by onnx.

Hello! It's a very meaningful work! I convert the mxnet model to onnx successfully and I can inference the visualization result by mxnet. I have three questions.
(1)Inference by onnxruntime. It got the following errors when I inference by onnxruntime.

(2) convert tensorrt error. It appears that the checkpoint.trt is not found when I convert mxnet to tensorrt.

(3) Densepose. I want to got the result of IUV array. How can I get the IUV array which is used to get the following result?

check failed

I wanted to start the Dense Pose: python3 launch.py --config configs/densepose_baseline/dense_pose_r50v1.py --task visualize --image imgs/single.json

and get the following error:
05-28 03:13:44 Visualization type: densepose
05-28 03:13:44 {'image': {'gt_bbox': [0, 0, 256, 512],
'im_id': 0,
'image_url': 'imgs/pitt_512_256.jpg'}}
Traceback (most recent call last):
File "launch.py", line 24, in
visualize_net(config=args.config, epoch=args.epoch, image=args.image)
File "/home/nano-ai/projects/poi/poi/apis/visualize.py", line 34, in visualize_net
pGen.batch_image, pIO.transform, pIO.data_name, None, logger)
File "/home/nano-ai/projects/poi/poi/infer/mx.py", line 12, in init
self.init_model(model_prefix, model_epoch, ctx_id)
File "/home/nano-ai/projects/poi/poi/infer/mx.py", line 16, in init_model
arg_params, aux_params = load_params(model_prefix, model_epoch)
File "/home/nano-ai/projects/poi/poi/core/model.py", line 8, in load_params
save_dict = nd.load("%s-%04d.params" % (prefix, epoch))
File "/usr/local/lib/python3.6/dist-packages/mxnet/ndarray/utils.py", line 175, in load
ctypes.byref(names)))
File "/usr/local/lib/python3.6/dist-packages/mxnet/base.py", line 246, in check_call
raise get_last_ffi_error()
mxnet.base.MXNetError: Traceback (most recent call last):
File "/home/nvidia/mxnet/mxnet/3rdparty/dmlc-core/src/io/local_filesys.cc", line 209
LocalFileSystem: Check failed: allow_null: :Open "logs/dense_pose_r50v1/checkpoint-0150.params": No such file or directory

any ideas?