-
As suggested, one class needs 2000 iterations. But the number of images may vary. Why we emphasize iterations rather than epoch?
-
### Search before asking
- [X] I have searched the YOLOv3 [issues](https://github.com/ultralytics/yolov3/issues) and [discussions](https://github.com/ultralytics/yolov3/discussions) and found no si…
-
**Describe the bug**
The ONNX [LSTM operator](https://github.com/onnx/onnx/blob/master/docs/Operators.md#attributes-26) states that the `activations` attribute is optional. The [LSTM backend implem…
samgd updated
5 years ago
-
Hi,
I tried a lot of things and can't find out how to do it. When training with COCO but only 7 classes using yolov2-tiny as config and weights, and change the classes to 7 and filters to 60, the e…
-
https://arxiv.org/abs/1802.10171
CVPR 2018
-
Hi,
I have been training yolo3 on the COCO+Pascal dataset (on a subset of 6 classes, out of the 80).
I tried on two different experiment setups:
a ) GTX 1070
b) Two GTX 1080 Ti's
Both the experim…
-
set activation level when execute bulk or get operation.
-
Hi,
I have seen your paper and you have done a splendid job on the class activation maps generating algorithm. Also, you have uploaded your codes for the sanity check, deletion and insertion, as wel…
-
I roughly skim-through the code base and searched for classes with `::Get` method. Some of these class objects are global singletons and have potential thread safety issues. I list the initial assessm…
-
> More precisely, we use activations from the last layer of neural network as speaker embeddings. We aggregate the sigmoid outputs by summing all outputs class-wise over the whole audio excerpt to obt…