32 lines
1.4 KiB
Markdown
32 lines
1.4 KiB
Markdown
# Oekk Model Zoo
|
|
|
|
A collection of pre-trained, state-of-the-art models in different formats with scripts to generate them and converter tools.
|
|
|
|
## Models
|
|
|
|
#### Vision
|
|
|
|
- CNN (Convolutional Neural Networks)
|
|
- Object Detection & Image Segmentation
|
|
- MNIST models
|
|
|
|
### CNN
|
|
|
|
| Model Class | Reference | Description |
|
|
|-------------|-----------|-------------|
|
|
| [U-Net]() | [Olaf Ronneberger et al.](https://arxiv.org/abs/1505.04597) | CNN with very good performance on very different biomedical segmentation applications.
|
|
|
|
#### Object Detection & Image Segmentation
|
|
|
|
|
|
#### MNIST models
|
|
|
|
This collection of models take 32x32x1 images (or variations) as input, then classifies the images into 10 number categories. Solvers of the MNIST dataset, frequently used as guinea pigs.
|
|
|
|
| Model Class | Reference | Description |
|
|
|-------------|-----------|-------------|
|
|
| [LeNet](https://coolneng.duckdns.org/gitea/oekk/models/src/branch/master/CNN/lenet/gen_scripts) | [Yann LeCun et al.](http://vision.stanford.edu/cs598_spring07/papers/Lecun98.pdf) | LeNet is one of the first CNN for image recognition. |
|
|
| [SqueezeNet](https://coolneng.duckdns.org/gitea/oekk/models/src/branch/master/CNN/squeezenet/gen_scripts)| [Iandola et al.](https://arxiv.org/abs/1602.07360) | A light-weight CNN model providing AlexNet level accuracy with 50x fewer parameters. Top-5 error from paper - ~20% |
|
|
|
|
|