# Image classification
Since I do not own enough computing power to iterate over ImageNet full training, this section involves training on a subset of ImageNet, called Imagenette.
## Getting started
Ensure that you have holocron installed
git clone https://github.com/frgfm/Holocron.git
pip install -e Holocron/. --upgradeDownload Imagenette and extract it where you want
wget https://s3.amazonaws.com/fast-ai-imageclas/imagenette2-320.tgz
tar -xvzf imagenette2-320.tgzFrom there, you can run your training with the following command
python train.py imagenette2-320/ --model darknet53 --lr 5e-3 -b 32 -j 16 --epochs 40 --opt radam --sched onecycle --loss label_smoothing
| Size (px) | Epochs | args | Top-1 accuracy | # Runs |
|---|---|---|---|---|
| 224 | 5 | imagenette2-320/ --model darknet53 --lr 5e-3 -b 32 -j 16 --epochs 5 --opt radam --sched onecycle --loss label_smoothing | 66.88% | 1 |
| 224 | 10 | imagenette2-320/ --model darknet53 --lr 5e-3 -b 32 -j 16 --epochs 10 --opt radam --sched onecycle --loss label_smoothing | 76.18% | 1 |
| 224 | 20 | imagenette2-320/ --model darknet19 --lr 5e-4 -b 32 -j 16 --epochs 20 --opt radam --sched onecycle --loss label_smoothing | 84.43% | 1 |
| 256 | 20 | imagenette2-320/ --model rexnet1_0x --lr 3e-4 -b 64 -j 8 --epochs 20 --opt tadam --sched onecycle --loss label_smoothing | 84.66% | 1 |
| 224 | 40 | imagenette2-320/ --model darknet19 --lr 5e-4 -b 32 -j 16 --epochs 40 --opt radam --sched onecycle --loss label_smoothing | 90.47% | 1 |
## Model zoo
| Model | Accuracy@1 (Err) | Param # | MACs | Interpolation | Image size |
|---|---|---|---|---|---|
| cspdarknet53 | 91.85 (8.15) | 26.63M | 5.03G | bilinear | 256 |
| rexnet2_2x | 91.75 (8.25) | 19.49M | 1.88G | bilinear | 224 |
| rexnet50d | 91.46 (8.54) | 23.55M | 4.35G | bilinear | 224 |
| darknet53 | 91.46 (8.54) | 40.60M | 9.31G | bilinear | 256 |
| repvgg_a2 | 91.26 (8.74) | 48.63M | bilinear | 224 | |
| darknet19 | 91.11 (8.89) | 19.83M | 2.75G | bilinear | 224 |
| tridentresnet50 | 91.01 (8.99) | 45.83M | 35.9G | bilinear | 224 |
| sknet50 | 90.42 (9.58) | 35.22M | 5.96G | bilinear | 224 |
| rexnet1_3x | 90.32 (9.68) | 7.56M | 0.68G | bilinear | 224 |
| repvgg_a1 | 90.19 (9.81) | 30.12M | bilinear | 224 | |
| rexnet1_0x | 90.01 (9.99) | 4.80M | 0.42G | bilinear | 224 |
| repvgg_a0 | 89.96 (9.04) | 24.74M | bilinear | 224 | |
| repvgg_b0 | 89.61 (9.39) | 31.85M | bilinear | 224 | |
| res2net50_26w_4s | 89.58 (99.26) | 23.67M | 4.28G | bilinear | 224 |
| darnet24 | 88.25 (11.75) | 22.40M | 4.21G | bilinear | 224 |
| resnet50 | 84.36 (15.64) | 23.53M | 4.11G | bilinear | 224 |