YOLO:darkflow [hard to train, fail to run demo]

Dependencies

Python3, tensorflow 1.0, numpy, opencv 3.

Source Code

Install

sudo apt-get install python3-numpy
sudo pip3 install .
python3 setup.py build_ext --inplace
#gedit darwflow/flow
change "#! /usr/bin/env python" to "#! /usr/bin/env python3"
source set-opencv3-python3.sh
./flow --h

YOLO:YOLOv1 tensorflow

Source Code

https://github.com/hizhangp/yolo_tensorflow

YOLO: YOLOv2 Chainer

Source Code

https://github.com/leetenki/YOLOv2

YOLO [tf-slim]

https://github.com/ruiminshen/yolo-tf

YAD2K [keras]

https://github.com/allanzelener/YAD2K

TensorBox [hard to train, fail to run demo]

Source Code

https://github.com/TensorBox/TensorBox

Faster-RCNN

PVANet [run demo failed!]

https://github.com/sanghoon/pva-faster-rcnn

py-faster-rcnn [run demo okay! but not good to train]

https://github.com/rbgirshick/py-faster-rcnn

donot use master version of caffe-fast-rcnn

 caffe-fast-rcnn git:(0dcd397)

donot use cudnn if you use cuda-8.0

# cuDNN acceleration switch (uncomment to build with cuDNN).
# USE_CUDNN := 1

donot pay attention to make test if error happend

src/caffe/test/test_smooth_L1_loss_layer.cpp:11:35:
fatal error: caffe/vision_layers.hpp: No such file or directory

donot pay attention to make pytest if error happend

AttributeError: 'SimpleParamLayer' object has no attribute 'param_str'
AttributeError: 'LayerParameter' object has no attribute 'shape'
FAILED (errors=4)
Makefile:518: recipe for target 'pytest' failed
make: *** [pytest] Error 1

cd lib and make

/home/yzbx/git/gnu/py-faster-rcnn/lib

Faster-RCNN [tf]

https://github.com/endernewton/tf-faster-rcnn

not good to train own dataset

Support for train-and-validationls
Support for resuming training
Support for visualization
Good For Small Object Detection For keeping the small proposals ( < 16 pixels width/height)

https://github.com/smallcorgi/Faster-RCNN_TF

not good to train own dataset.

This is an experimental Tensorflow implementation of Faster RCNN - a convnet for object detection with a region proposal network.

https://github.com/CharlesShang/TFFRCNN

Training on KITTI detection dataset
Hard Example Mining
Data Augment
PVANet

requirement

1. Update your -arch in setup script to match your GPU

cd tf-faster-rcnn/lib
# Change the GPU architecture (-arch) if necessary
vim setup.py

GPU Model	Architecture
TitanX (Maxwell/Pascal)	sm_52
GTX 960M	sm_50
GTX 1080 (Ti)	sm_61
Grid K520 (AWS g2.2xlarge)	sm_30
Tesla K80 (AWS p2.xlarge)	sm_37

2. Set environment to tensorflow-gpu=1.0

conda install tensorflow-gpu=1.0
conda install cython matplotlib scipy Pillow pyyaml tqdm scikit-image
pip install easydict opencv_python

3. Set gcc for tensorflow

tensorflow.python.framework.errors_impl.NotFoundError: 
TFFRCNN/lib/roi_pooling_layer/roi_pooling.so: 
undefined symbol: _ZN10tensorflow7strings6StrCatB5cxx11ERKNS0_8AlphaNumE

vim lib/make.sh
## if you install tf using already-built binary, or gcc version 4.x, uncomment the two lines below
... D_GLIBCXX_USE_CXX11_ABI=1 ...
# ... D_GLIBCXX_USE_CXX11_ABI=0 ...

4. Set --restore 0

AttributeError: 'NoneType' object has no attribute 'model_checkpoint_path'

python ./faster_rcnn/train_net.py --gpu 0 --weights /media/sdb/VGG_imagenet.npy
 --imdb voc_2007_trainval --iters 70000 --restore 0 --cfg  ./experiments/cfgs/faster_rcnn_end2end.yml 
 --network VGGnet_train --set EXP_DIR exp_dir