SKU110K-benchmark

Evaluation code for SKU110K dataset. Forked from cocoapi repository.

Steps

• Place the annotations_test.csv inside groundtruth folder. Data can be obtained from here [1].
• Place your predictions.csv inside predictions folder.
• Run calc_acc.py inside cocoapi/PythonAPI/pycocotools after modifying the predictions file path.

Results

Method	AP	AP.75	AR300
Goldman et al [2]	0.492	0.556	0.554
Our method [3]	0.685	0.755	0.756

Note

The outputs from [3] can be downloaded from this link

References

[1] https://github.com/eg4000/SKU110K_CVPR19
[2] https://arxiv.org/abs/1904.00853
[3] Will be released soon