tensorflow / tensorflow/models
SSD MobilenetV1 on COCO - pre-trained protobuf and checkpoint mAP don't match
Open
@pkulzc is already working on this.
Since Jun 22, 2020.
models:research:odapi
type:support
- Dominant language
- Python
- Stars
- 77.7k
- Forks
- 44.8k
- PR merge metrics
- No merged PRs in 30d
Description
Using the pre-trained files, frozen_inference_graph.pb is giving an mAP of 18% va checkpoint mAP of 23%.
I have tried debugging, but I am unable to find the issue.
The only difference between the evaluation for 2 cases is- For checkpoints, the input is a tfrecord file whereas for .pb case, the inputs are .jps.
The image list is the same for both.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.