tensorflow / tensorflow/models
Number of anchor boxes / priors don't match with original SSD paper
Open
@pkulzc is already working on this.
Since May 20, 2020.
models:research:odapi
type:support
- Dominant language
- Python
- Stars
- 77.7k
- Forks
- 44.8k
- PR merge metrics
- No merged PRs in 30d
Description
Looking at the calculations for number of priors for SSD as mentioned in the original paper, it should be 8732 priors.
But based on the below test cases, we are only getting 7308 priors.
https://github.com/tensorflow/models/blob/master/research/object_detection/anchor_generators/multiple_grid_anchor_generator_test.py#L271
They seem to match for the middle 3 layers, but not for layer 1, 5 and 6.
Observations
| Layer | Expected based on SSD paper | Observed in model framework | Difference |
|---|---|---|---|
| (38, 38) | 5776 | 4332 | 1444 |
| (19, 19) | 2166 | 2166 | 0 |
| (10, 10) | 600 | 600 | 0 |
| (5, 5) | 150 | 150 | 0 |
| (3, 3) | 36 | 54 | -18 |
| (1, 1) | 4 | 6 | -2 |
| Total missing | 8732 | 7308 | 1424 |
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.