Toru Ogawa
Toru Ogawa
I think that issue should be fixed in the ImageNet example.
tests for `utils/test.py` is added in #141. (`test.py` is renamed in that PR)
Please try `model_*.npz` instead of `snapshot_*.npz`. snapshots contain additional information for resuming and their structure is different from that of models.
Could you share the version of ChainerCV and the commit id of the training code?
If using `cv2.imread` is fastest, we can simply use `cv2.imread` in `read_image`. This doesn't require any changes of APIs. If you want to use `PIL`, we have to change APIs.
> I am guessing that most of the time is spent decoding jpg image, and it seems that cv2 has a better decoder. I guess this is depends on the...
In my environment, cv2 was fastest, too. ``` crop_cv2 mean=0.22242753505706786 crop_pil mean=0.34299739996592205 crop_numpy mean=0.38132399717966714 ```
I think it is better to unify keywords. I guessed current rules from the codes.
@yuyu2172 How do think of it? Do you know other implicit rules?
@yuyu2172 thank you. I merged the rules you suggested. I changed the state of usage of `class` to "under discussion".