This method is inspired by FCOS: Fully Convolutional One-Stage Object Detection, the implementation is derived from DetectionTeamUCAS.
1、python3.5
2、cuda8.0
3、opencv(cv2)
4、tfplot
5、tensorflow >= 1.12
6、MATLAB / GNU Octave
1、Please download resnet50_v1, resnet101_v1 pre-trained models on Imagenet, put it to data/pretrained_weights.
1、Please download kaist_model , and put it in output/trained_weights
2、The .tfrecord must placed in data/tfrecord
3、Download the results images and text files , and put it in output/test_results
cd $PATH_ROOT/libs/box_utils/cython_utils
python setup.py build_ext --inplace
cd $PATH_ROOT/libs/box_utils/nms
python setup.py build_ext --inplace
1、Train
Make sure pretrained models and tfrecord are in repsective directories
cd $PATH_ROOT/tools
python multi_gpu_train_v2.py --GPU 1
1、Testing on the KAIST - Test Set
Make sure trained models is in the correct directory
cd $PATH_ROOT/tools
python test.py --rgb_data_dir PATH_TO/Reasonable_test_depth/visible --ir_data_dir PATH_TO/Reasonable_test_depth/lwir --save_dir PATH_TO/output/test_results --GPU 0
1、Evaluating the miss-rate of the obtained results
2、Copy results from output/test_results/txt -> eval/det
3、Arrange Ground Truth annotations in the following format
Ground_truth_directory
--test-all
----annotations (these are orignal annotations)
----annotations_KAIST_test_set (these are imporved annotations)
--test-day
----annotations (these are orignal annotations)
----annotations_KAIST_test_set (these are imporved annotations)
--test-night
----annotations (these are orignal annotations)
----annotations_KAIST_test_set (these are imporved annotations)
4、Evaluate Metrics
run demo_test.m from MATLAB/GNU OCTAVE
1、https://github.com/endernewton/tf-faster-rcnn
2、https://github.com/zengarden/light_head_rcnn
3、https://github.com/tensorflow/models/tree/master/research/object_detection
4、https://github.com/CharlesShang/FastMaskRCNN
5、https://github.com/matterport/Mask_RCNN
6、https://github.com/msracver/Deformable-ConvNets
7、https://github.com/tianzhi0549/FCOS
8、https://github.com/Li-Chengyang/MSDS-RCNN