hust-kevin

colab run error

1

@kwea123 I use colab to run colmap, but meet this error  can you tell me how to solve it

the time test on coco is much larger than voc

2

when i use the RFBNet300_VOC_80_7.pth to eval voc images(4952)，the total test time is 190s，but when i use the RFB_vgg_COCO_30.3.pth to eval minival2014 (5000 images)，the total test time is 1000s, why...

why calculate balanced_weights use 1/log(1+x)

3

why calculate balanced_weights [here](https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/utils/calculate_weights.py#L23) use log `class_weight = 1 / (np.log(1.02 + (frequency / total_frequency)))` why not just set `class_weight = frequency / total_frequency`

why not add gn

2

sn learn importantance among bn，In，Ln，why not add gn

what's the mean of traing_mask?

1

hi, can you tell me what's the [traning_mask](https://github.com/whai362/PSENet/blob/master/dataset/ctw1500_loader.py#L204)?

Metric issues

1

Hello when I read 'metric.py'，I have a question def update(self, labels, preds): # output: cls_prob_output, bbox_pred_output, cls_keep_inds, bbox_keep_inds # label: label, bbox_target what do 'cls_keep_inds, bbox_keep_inds' mean ? Looking forward...

cascade rcnn真的增加了正样本的数量吗

1

我将rpn阶段下采样得到的512个proposal中的正负样本数量打印出来，发现大多数情况下这512个proposal中也只有几个是正样本，似乎并没有增加正样本的数量呀

I try to use optimum convert nllb to onnx, but get error ``` onnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError] : 1 : FAIL : Deserialize tensor onnx::MatMul_3635 failed.tensorprotoutils.cc:637 TensorProtoToTensor External initializer: onnx::MatMul_3635 offset: 0...