`detection_features` for `ssd_*`

[malicd]

I am trying to extend SSDMetaArch such that postprocess(...) returns detection_features (same as faster_rcnn_*). Is there a way to determine which box originates from which feature map, i.e. to map each box_encodings with feature_maps from raw prediction_dict?

[tensorflowbutler]

Thank you for your post. We noticed you have not filled out the following field in the issue template. Could you update them if they are relevant in your case, or leave them as N/A? Thanks.
What is the top-level directory of the model you are using
Have I written custom code
OS Platform and Distribution
TensorFlow installed from
TensorFlow version
Bazel version
CUDA/cuDNN version
GPU model and memory
Exact command to reproduce

[malicd]

What is the top-level directory of the model you are using: models/research
Have I written custom code: Yes
OS Platform and Distribution: Ubuntu 18.04
TensorFlow installed from: Binary (pip)
TensorFlow version: 1.13.1
Bazel version: /
CUDA/cuDNN version: 10.0/7.4.2
GPU model and memory: GTX 1080Ti (11GB)
Exact command to reproduce: /

[qraleq]

@malicd Do you have any update on this issue?

[malicd]

This was a lot harder then I anticipated since the feature maps come at different shapes and depths. It was out of the scope for my needs and I end up using Faster RCNN (it is already implemented there).