First attempt - switched to tensorflow object_detection. See the motion-rcnn repo.
- tensorflow (>= 1.2.0) with GPU support. For best performance, i highly recommend building from source.
pip install pillow matplotlib opencv-python easydict cython tqdm
- create
./outdirectory - copy
env_template/env.ymltoout/env.ymland adapt for your machine setup - download
http://download.tensorflow.org/models/resnet_v1_50_2016_08_28.tar.gzand unzip to./data/models/ - go to
./liband runmake - run
tools/create_tfrecords.pywith each--dataset/--splitcombination you need
- run
python tools/trainval.pyfor training - run
python tools/test.pyfor testing
- run
python test/cityscapes.pyto visualize the cityscapes dataset - run
python test/anchors.pyto visualize anchors for different levels
Visualizations are written to out/tests.
- The code in
lib/nmsandlib/boxesis taken without changes from py-faster-rcnn. - The tensorflow code in
lib/nets/resnet_v1.pyandlib/nets/network.pyis based on tf-faster-rcnn. - The files implementing common Faster R-CNN layers in
lib/layersare based on py-faster-rcnn and include small modifications from tf-faster-rcnn. - The code in
lib/datasets/cityscapes/cityscapesscriptsis adapted from cityscapesScripts.
See LICENSE for details.