semantic error on the evaluation code

[realjoshqsun]

There's a semantic error in "_interpolate_recall_at_fp". Line 336-340: fp_0 = fp_recall_less_fp["fp"].values[-1] fp_1 = fp_recall_more_fp["fp"].values[0] recall_0 = fp_recall_less_fp["recall"].values[-1] recall_1 = fp_recall_more_fp["recall"].values[0]

The correct code should be: fp_0 = fp_recall_less_fp["fp"].values[0] fp_1 = fp_recall_more_fp["fp"].values[-1] recall_0 = fp_recall_less_fp["recall"].values[0] recall_1 = fp_recall_more_fp["recall"].values[-1]

[kaimingkuang]

Thanks for spotting this bug. A new commit has been pushed and now the FP-recall data frame is sorted by FP in ascending order first before calculating key recall since this is easier.