pretrained-models.pytorch
pretrained-models.pytorch copied to clipboard
Cadene
Pretrained ConvNets for pytorch: NASNet, ResNeXt, ResNet, InceptionV4, InceptionResnetV2, Xception, DPN, etc.
Hi! I'm trying to download pretrained densenet169 and getting error "certificate verify failed: certificate has expired". ``` import pretrainedmodels pretrainedmodels.densenet169(num_classes=1000, pretrained='imagenet') ``` ``` Downloading: "http://data.lip6.fr/cadene/pretrainedmodels/densenet169-f470b90a4.pth" to /root/.cache/torch/hub/checkpoints/densenet169-f470b90a4.pth Traceback (most recent...
I tried to load pre-trained model "inceptionresnetv2", but it failed to connect. The command as below: pretrainedmodels.__dict__[inceptionresnetv2](num_classes=1000, pretrained='imagenet') and I got the error: HTTPConnectionPool(host='data.lip6.fr', port=80): Max retries exceeded with url:...
When running pytest with the latest version of Pillow, you'll see: ``` ../../../anaconda/envs/adam-py37-min/lib/python3.7/site-packages/pretrainedmodels/datasets/utils.py:33 /anaconda/envs/adam-py37-min/lib/python3.7/site-packages/pretrainedmodels/datasets/utils.py:33: DeprecationWarning: BILINEAR is deprecated and will be removed in Pillow 10 (2023-07-01). Use Resampling.BILINEAR instead. def...
Hello, I changed weight initialization process of xception model as the user had to pass the same `num_classes` within `pretrained_settings`.
Inception resenet v2 stem replicates the stem of v1 instead of v2 as described in the original paper
The stem implemented in inception resenet v2 has replicated the stem of inception resnet v1. The two are however different in the original paper [https://arxiv.org/pdf/1602.07261.pdf](url)
Take Xception models as an example. There are three weight files-"xception-43020ad28.pth","xception-b0b7af25.pth","xception-b5690688.pth" in [link](https://data.lip6.fr/cadene/pretrainedmodels/). What are the differences?
https://github.com/Cadene/pretrained-models.pytorch/blob/8aae3d8f1135b6b13fed79c1d431e3449fdbf6e0/pretrainedmodels/models/inceptionresnetv2.py#L300 In this realization model has 1536 feature maps in the end. However, according to https://arxiv.org/pdf/1602.07261.pdf it should be 2048 (fig 19). Please, can you explain, why the realization is...
Would it be possible to have also greyscale ImageNet weights for the models along the lines described in Xie & Richmond? > Xie, Y. and Richmond, D., “[Pre-training on grayscale...
When I import the pretrained models, by default, it downloads at first in `/home/user/.cache/torch/hub/ ..model*` Is it possible to set this to any other directory? Please if you can guide...
I am getting an SSL Error trying to load ImageNet weights for the SEResNeXt50 & 101 models. Perhaps there is an expired certificate? Here is code to reproduce. Thanks in...
