keysmis / tf-pose-estimation

Openpose from CMU implemented using Tensorflow with Custom Architecture for fast inference.

tf-pose-estimation

'Openpose' for human pose estimation have been implemented using Tensorflow. It also provides several variants that have made some changes to the network structure for real-time processing on the CPU or low-power embedded devices.

Features in this fork

• Add support for using pretrained model : mobilenet_V2, VGG16x4 (pruned VGG16 model, 4 times faster)
• trainging with multi-scale loss
• more mobilenet networks config
• bug fix in the original repo
• evalution script on validation dataset (support evaluation of both tf and caffe model)
• tensorrt support

Inference From Caffemodel (cmu openpose)

python3 src/inference_cmupose.py --input-width=656

Inference From Mobilenet_thin

python3 src/inference.py --input-width=656

Inference From Tensorrt engine

python3 src/tensorrt_inference.py --graph=yourgraph.opt.pb --engine=yourmodel.engine

Inference and run from webcam With Tensorflow with tensorrt built-in

python3 src/run_webcam.py --use_tensorrt=1