Hi, thanks for the great code!

I wonder do you have plans to support resuming from checkpoints for classification? As we all know, in terms of training ImageNet, the training process is really long and it can be interrupted somehow, but I haven't notice any code related to "resume" in scripts/classification/train_imagenet.py.

Maybe @hetong007 ? Thanks in advance.

My actions before raising this issue

• Read/searched the docs
• Searched past issues

Current version of pyAV is 6.2.0. Need to update it till 8.x.
Also PyAV d

Description

Consider this very short piece of code:

testimg = np.array([0, 1, 2, 3, 4, 3.5, 4, 5, 4, 3, 2, 1, 0])
morphology.h_maxima(testimg, 1)
morphology.h_maxima(testimg, 2)
morphology.h_maxima(testimg, 3)

The results show that the element with value 5 is always picked up by this function, even though it is not a local maximum with height h for any h; rather it i

I'm new to tracking, and I have a question about the test data.
In test, OTB2015 VOT16/17/18 are supported, but in train, COCO DET VID are supported.
What's the connection between these?
If I want to run train, what data should I use?
Test?
Thank you!

(siammask) [liqiang@inspur siammask]$ bash test_mask_refine.sh config_vot.json SiamMask_VOT.pth VOT2016 0
[2019-03-14 19:42:16,619-rk0-test.py#551] Namespace(arch='Custom', config='config_vot.json', dataset='VOT2016', gt=False, log='log_test.txt', mask=True, refine=True, resume='SiamMask_VOT.pth', save_mask=False, visualization=False)
[2019-03-14 19:42:17,087-rk0-load_helper.py# 31] load pretrai

Link: https://kornia.readthedocs.io/en/latest/geometry.conversions.html#kornia.geometry.conversions.angle_axis_to_quaternion

Input and output are described as:
Input: (∗,3) where * means, any number of dimensions
Output: (∗,4)

While the example shows:

angle_axis = torch.rand(2, 4) # Nx4 quaternion = kornia.angle_axis_to_quaternion(angle_axis) # Nx3

The example indicates that in

Feature Request

I want to obtain the evaluation of different datasets of a single job
lumi eval -h && lumi eval -c config_ssd.yml --split valid --split train --split test --watch --from-global-step 0

I understand that it is not so trivial because it was being kept as a summary of work scalars like this
![image](https://user-images.githubusercontent.com/7362688/48442210-8be6d080-e75b