ssd.pytorch

ssd.pytorch copied to clipboard

RuntimeError: stack expects each tensor to be equal size, but got [1, 5] at entry 0 and [2, 5] at entry 1

1

two different images, the first image has one object, and the second image has two object ``` def fetch(self, possibly_batched_index): if self.auto_collation: data = [self.dataset[idx] for idx in possibly_batched_index] else:...

cs-heibao

Traceback (most recent call last): File "test.py", line 101, in test_voc() File "test.py", line 98, in test_voc thresh=args.visual_threshold) File "test.py", line 49, in test_net img_id, annotation = testset.pull_anno(i) ValueError: too...

Zhezhefufu

RuntimeError: cuda runtime error (11) :invalid argument at /pytorch/aten/src/THC/THCGeneral.cpp:383

2

OS: Ubuntu python: 3.6 cuda:9.0 pytorch:1.1.0(GPU) GPU: RTX2080(8g) When I execute the command: `python3 train.py --dataset VOC --dataset_root /media/gaoya/disk/Datasets/VOCdevkit/ --basenet vgg16_reducedfc.pth --batch_size 8` **Error happened:** Loading base network... Initializing weights......

ouening

ONNX export compatible with OpenCV

18

This PR has changes which let to export trained model to ONNX format compatible with OpenCV (PR https://github.com/opencv/opencv/pull/16925 is required) ### Convert to ONNX ``` python3 export_to_onnx.py --model ssd300_mAP_77.43_v2.pth ```...

dkurt

fixed number of classes for coco in config file

Nannigalaxy

Fix COCO configuration

14

I tried the training and testing on COCO dataset and fixed some minor bugs in the configuration.

chienyiwang

jetson_nano@Jetson:~/jetson-inference/python/training/detection/ssd$ python3 onnx_export.py --model-dir=models/markusmodel Namespace(batch_size=1, height=300, input='', labels='labels.txt', model_dir='models/markusmodel', net='ssd-mobilenet', output='', width=300) running on device cuda:0 found best checkpoint with loss 0.760933 (models/markusmodel/mb1-ssd-Epoch-29-Loss-0.7609327760609713.pth) creating network: ssd-mobilenet num classes: 4 loading...

MST2010Markus

I trained my own dataset. But loss didn't decrease instead increase

13

My datasets has 5 classes. batch size = 8, lr=1e-5. others config just the same as orginal code. **Here is my trained progress!**     Is there a...

tangdong1994

batch evalutation

1

Has someone been able to modify the evaluation script to allow for the use of batches to make it (a lot) faster on GPU?

JappaB

Loss is about 2.x, but mAP is not good, only about 66.x%.

4

I have tried many training tricks, such as different batch_size, different learning rate. But, almost each loss is 2.x. I checked other issues, their mAP are 7x.x%, but mine is...

XiangqianMa