-
I try to execute the command " python hdc_test.py -d cars196 -c HDC" using the model trained by you. But
in the script hdc_test.py, when choose HDC method, the function test_cars.eval is called. S…
-
I use the same triplet loss (with BatchHard, Euclidean distance and Soft-margin) on the fine-grained categorization dataset [CUB-200-2011.](http://www.vision.caltech.edu/visipedia/CUB-200-2011.html) …
-
안녕하세요.
Clova vision팀의 고병수입니다.
제가 오프라인 결선에서 Clova vision팀의 소개자료 중에 저희팀이 연구하고 있는 논문을 공유드린다고 말씀드린게 있었는데, 이제서야 arxiv에 공개가 되었습니다. 🎉🎉🎉
논문은 image retrieval에 최적화된 framework를 제안하고 있고, 대표적인 데이터셋인 CARS196, …
-
이상하게 mtcnn을 써서 bounding box로 cropping한 이미지로 학습/평가를 하면 성능이 악화되는듯한 느낌?
이미지 사이즈와 resolution의 영향인가?
- 이미지 사이즈를 112 x 96 으로 바꿔보기. (최신 논문들은 보통 이 사이즈로 실험을 한다)
- Image Net으로 pretrain된 모델 사용하지 말고 random …
-
Hi there!
Thank you for the awesome library!
I'm currently working on training a model using the CARS196 dataset with the following parameters:
```python
distance = distances.CosineSimilarit…
-
Datasets such as CIFAR / MNIST / etc have download / integrity logic that is useful for reimplementing custom user datasets (such as https://github.com/vadimkantorov/metriclearningbench/blob/master/ca…
-
some codes in `train.py` is like:
```python
if world_size > 1:
with torch.no_grad():
all_z = [torch.zeros_like(z) for _ in range(world_size)]
torch.distributed.all_gath…
-
Hi, @bnu-wangxun
I want to reproduce the result of Binomial Loss, so I modified the run_train_00.sh slightly:
```
DATA=car
...
LOSS=Binomial
...
DIM=64
```
And I downloaded the bninception f…
-
Hi, I think its important to provide scripts to reproduce paper results for the models the package implements. To be some kind of proof everything is working correctly.
What do you think?
Im not sur…
-
### Description
The current SOTA paper for image retrieval was published in 2020. It's called "Combination of Multiple Global Descriptors for Image Retrieval," and you can find it [here](https://arxi…