add third-party implementation, fix data setup instruction, and fix typo in dlav0

README.md

@@ -119,6 +119,11 @@ We provide scripts for all the experiments in the [experiments](experiments) fol
 
 If you are interested in training CenterNet in a new dataset, use CenterNet in a new task, or use a new network architecture for CenterNet, please refer to [DEVELOP.md](readme/DEVELOP.md). Also feel free to send us emails for discussions or suggestions.
 
+## Third-party implementation
+
+- Keras: [keras-centernet](https://github.com/see--/keras-centernet) from [see--](https://github.com/see--).
+
+
 ## License
 
 CenterNet itself is released under the MIT License (refer to the LICENSE file for details).

readme/DATA.md

@@ -91,6 +91,13 @@ If you want to reproduce the results in the paper for benchmark evaluation and t
 
 - Run `python convert_kitti_to_coco.py` in `tools` to convert the annotation into COCO format. You can set `DEBUG=True` in `line 5` to visualize the annotation.
 
+- Link image folder
+
+  ~~~
+  cd ${CenterNet_ROOT}/data/kitti/
+  mkdir images
+  ln -s training/image_2 images/trainval
+  ~~~
 
 - The data structure should look like:
 

src/lib/models/networks/dlav0.py

@@ -639,8 +639,8 @@ def dla169up(classes, pretrained_base=None, **kwargs):
     return model
 '''
 
-def get_pose_net(heads, down_ratio=4, head_conv=256):
+def get_pose_net(num_layers, heads, add_conv=256, down_ratio=4):
-  model = DLASeg('dla34', heads,
+  model = DLASeg('dla{}'.format(num_layers), heads,
                  pretrained=True,
                  down_ratio=down_ratio,
                  head_conv=head_conv)