google-deepmind / google-deepmind/kinetics-i3d
Training with different architectures
Open
- Dominant language
- Python
- Stars
- 1.8k
- Forks
- 467
- PR merge metrics
- No merged PRs in 30d
Description
Hi, has anyone tried inflating a model pretrained on Imagenet from 2D to 3D and training on Kinetics using models different from Inception-v1 (like Resnets or Densenets)?. If you have code that you can share I would be very greatful.
Contributor guide
Research direction
The issue does not name a file, test, or implementation entry point; it asks whether 2D ImageNet-pretrained architectures such as ResNet or DenseNet have been inflated for 3D Kinetics training. First review the repository's model and training entry points, then establish a concrete architecture target and success criteria before work can be scoped.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- computer-vision, machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 15/100