|Submission name||Zhigang-CDPNv2 (MODE 2, FCOS)|
|Submission time (UTC)||Aug. 19, 2020, 6:45 a.m.|
|Task||6D localization of seen objects|
|Training model type||Default|
|Training image type||Synthetic (only PBR images provided for BOP Challenge 2020 were used)|
|Publication||CDPN: Coordinates-Based Disentangled Pose Network for Real-Time RGB-Based 6-DoF Object Pose Estimation|
|Training image modalities||RGB|
|Test image modalities||RGB|
In PBR-only setting, all models are trained only using the provided PBR synthetic data. For each dataset, we trained a CDPN model for each object.
For detection, different from CDPN in BOP19, we used the FCOS with BackBone of vovnet-V2-57-FPN . We trained a detector for each dataset. The detector was trained for 8 epochs with batch size of 4 on a single GPU, 4 workers, and a learning rate of 1e-3. We used color augmentation similar to AAE  during training.
For pose estimation, the difference between our CDPNv2 and the BOP19-version CDPN mainly including:
Besides the color augmentation similar to AAE , we also used the truncation domain randomization in  to improve the system robustness to occlusion.
Considering the organizer provides high-quality PBR synthetic training data in BOP20, we adopt a deeper 34-layer Resnet as the backbone instead of the 18-layer Resnet used in BOP19-version CDPN. Also, the fancy concat structures in BOP19-version CDPN are removed. The input and output resolutions are 256×256 and 64×64 respectively.
During training, the initial learning rate was 1 × 10−4 and the batch size was 6. We used RMSProp with alpha 0.99 and epsilon 1× 10−8 to optimize the network. The model was trained for 160 epochs in total and the learning rate was divided by 10 every 50 epochs
|Computer specifications||Intel i7-7700; GPU: GTX 1070; Memory: 16G|