mirrors/pytorch-image-models
1
0
Fork 0
mirror of https://github.com/huggingface/pytorch-image-models.git synced 2025-06-03 15:01:08 +08:00
Code Issues Projects Releases Wiki Activity

PyTorch trained MobileNetV3 weights that match/best paper. Update fbnetc weights to improved copy.

Browse Source
This commit is contained in:
Ross Wightman 2019-06-04 09:51:05 -07:00
parent 89e77ec049
commit 6688d0669f
2 changed files with 13 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
README.md
View File

@ -62,12 +62,14 @@ Several (less common) features that I often utilize in my projects are included.
### Self-trained Weights
I've leveraged the training scripts in this repository to train a few of the models with missing weights to good levels of performance. These numbers are all for 224x224 training and validation image sizing with the usual 87.5% validation crop.
#### @ 224x224
|Model | Prec@1 (Err) | Prec@5 (Err) | Param # | Image Scaling |
|---|---|---|---|---|
| resnext50_32x4d | 78.512 (21.488) | 94.042 (5.958) | 25M | bicubic |
| seresnext26_32x4d | 77.104 (22.896) | 93.316 (6.684) | 16.8M | bicubic |
| mobilenetv3_100 | 75.634 (24.366) | 92.708 (7.292) | 5.5M | bicubic |
| fbnetc_100 | 75.124 (24.876) | 92.386 (7.614) | 5.6M | bilinear |
| resnet34 | 75.110 (24.890) | 92.284 (7.716) | 22M | bilinear |
| fbnetc_100 | 74.830 (25.170 | 92.124 (7.876) | 5.6M | bilinear |
| seresnet34 | 74.808 (25.192) | 92.124 (7.876) | 22M | bilinear |
| spnasnet_100 | 74.084 (25.916) | 91.818 (8.182) | 4.42M | bilinear |
| seresnet18 | 71.742 (28.258) | 90.334 (9.666) | 11.8M | bicubic |

13
models/gen_efficientnet.py
View File

@ -65,11 +65,15 @@ default_cfgs = {
'mobilenetv2_100': _cfg(url=''),
'mobilenetv3_050': _cfg(url=''),
'mobilenetv3_075': _cfg(url=''),
'mobilenetv3_100': _cfg(url=''),
'mobilenetv3_100': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/mobilenetv3_100-35495452.pth',
interpolation='bicubic'),
'chamnetv1_100': _cfg(url=''),
'chamnetv2_100': _cfg(url=''),
'fbnetc_100': _cfg(url='https://www.dropbox.com/s/0ku2tztuibrynld/fbnetc_100-f49a0c5f.pth?dl=1'),
'spnasnet_100': _cfg(url='https://www.dropbox.com/s/iieopt18rytkgaa/spnasnet_100-048bc3f4.pth?dl=1'),
'fbnetc_100': _cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/fbnetc_100-c345b898.pth'),
'spnasnet_100': _cfg(
url='https://www.dropbox.com/s/iieopt18rytkgaa/spnasnet_100-048bc3f4.pth?dl=1'),
'efficientnet_b0': _cfg(url=''),
'efficientnet_b1': _cfg(url='', input_size=(3, 240, 240)),
'efficientnet_b2': _cfg(url='', input_size=(3, 260, 260)),
@ -1335,6 +1339,9 @@ def mobilenetv3_075(num_classes, in_chans=3, pretrained=False, **kwargs):
def mobilenetv3_100(num_classes, in_chans=3, pretrained=False, **kwargs):
""" MobileNet V3 """
default_cfg = default_cfgs['mobilenetv3_100']
if pretrained:
# pretrained model trained with non-default BN epsilon
kwargs['bn_eps'] = _BN_EPS_TF_DEFAULT
model = _gen_mobilenet_v3(1.0, num_classes=num_classes, in_chans=in_chans, **kwargs)
model.default_cfg = default_cfg
if pretrained: