search

ikuinen / CMIN_moment_retrieval

Cross-Modal Interaction Networks for Query-Based Moment Retrieval in Videos
86 stars 20 forks source link

About the results #3

Closed wangsu1999 closed 5 years ago

wangsu1999 commented 5 years ago

Hi,

I couldn't reach the results of your papers,and the gap was two to three percentage points.Could you please check the codes ?

Thanks, Looking forward to your reply!

ikuinen commented 5 years ago

Hi I ran the code again on the ActivityCaptions dataset, and still achieved the results even better than the results given by the paper. The evaluation is performed on the test set. Here is the evaluation results after four epochs of training: | loss, 0.2260 | mIoU, 0.4372 | IoU@0.1, 0.7938 | IoU@0.3, 0.6310 | IoU@0.5, 0.4308 | IoU@0.7, 0.2476 | IoU@0.9, 0.0964 |

wangsu1999 commented 5 years ago

Hi Thank you ! I tried to increase the learning rate in the model.I used 3e-3 instead of 8e-4, which made me get the results given by the paper.