my results are lower than what is indicated

[sofsoo1995]

Congratulation for your COCO Challenge result, and thank you for sharing your code.

I'm testing your code and the problem is I have these results on the validation set of 2014:

Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets= 20 ] = 0.430 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets= 20 ] = 0.682 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets= 20 ] = 0.458 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.385 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.511 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 20 ] = 0.534 Average Recall (AR) @[ IoU=0.50 | area= all | maxDets= 20 ] = 0.780 Average Recall (AR) @[ IoU=0.75 | area= all | maxDets= 20 ] = 0.566 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.463 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.631

-I'm using ResNet-101 with an input size of 384x288 -I haven't almost changed anything in the code(except the config file, in mptest I put dump_method=1 as argument for the function MultiProc). -I'm using the pretrained model you have trained. -Also, for the dataset, I have downloaded the 2014 version(train and val). -I'm using the annotation file and also the bounding box you've given. -I don't understand why is there a huge difference between your results and mine ! Have I done something wrong ?

[chenyilun95]

I've downloaded the pretrained model from web and ran it again using 8 gpus, got 72.9. Have you tried other pretrained model? Did they get the right result?

[sofsoo1995]

Thx you for your answers, The results I'm reporting is with the pretrained model and not my training. I only have tried res101.384x288 and res50.256x192 and the results are approximately the same(I've just reported res101.384x288). qualitatively, the model works on simple cases but it does a lot of errors for more complicated cases. I will try the other models and will also try to train(with 2 Titan X and 1 GTX 1080 )

[chenyilun95]

Do you check the validation dataset?
minival dataset: https://drive.google.com/drive/folders/15loPFQCMQnJqLK1viSMeIwTFT-KbNzdG minival det: https://drive.google.com/drive/folders/1BllF9--dN9uV3FRROcmuIbwNCcn7cCP0?usp=sharing

[sofsoo1995]

Ok, I've re downloaded pre trained model and I've re tested everything. and actually res50.256x192 has good results (as it is indicated) ! However the other methods failed (res50.384x288, and res101.384x288). So the problem might come from the input size. Is it possible that it is a problem of hardware(because I only use 3 GPUs) ?

I put the details here(with pretrained model):

res50.256x192

Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets= 20 ] = 0.697 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets= 20 ] = 0.883 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets= 20 ] = 0.770 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.662 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.761 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 20 ] = 0.764 Average Recall (AR) @[ IoU=0.50 | area= all | maxDets= 20 ] = 0.927 Average Recall (AR) @[ IoU=0.75 | area= all | maxDets= 20 ] = 0.823 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.715 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.830

res50.384x288

Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets= 20 ] = 0.206 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets= 20 ] = 0.459 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets= 20 ] = 0.162 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.170 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.268 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 20 ] = 0.310 Average Recall (AR) @[ IoU=0.50 | area= all | maxDets= 20 ] = 0.591 Average Recall (AR) @[ IoU=0.75 | area= all | maxDets= 20 ] = 0.280 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.250 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.392

res101.384x288

Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets= 20 ] = 0.430 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets= 20 ] = 0.682 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets= 20 ] = 0.458 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.385 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.511 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 20 ] = 0.534 Average Recall (AR) @[ IoU=0.50 | area= all | maxDets= 20 ] = 0.780 Average Recall (AR) @[ IoU=0.75 | area= all | maxDets= 20 ] = 0.566 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.463 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.631

[chenyilun95]

I think it can't be the problem of hardware. Since it's worse in 384x288 size, did you run the pre-trained model in the corresponding model folder?

[sofsoo1995]

I found the mistake. I modified config and with silly copy-pastes I put the wrong i input size in the two networks. My Bad !

However, I still don't understand why is there this huge difference of results if I put a wrong input size for testing.

But thank you very much for helping me !