rezapourian commited on
Commit
84347c8
·
verified ·
1 Parent(s): e70f653

Upload folder using huggingface_hub

Browse files
baselines/AMAES_original_README.md ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-nc-sa-4.0
3
+ datasets:
4
+ - FOMO-MRI/FOMO300K
5
+ pipeline_tag: image-segmentation
6
+ ---
7
+
8
+ # ResEnc UNet pretrained on FOMO300K using AMAES
9
+
10
+ This repository contains the checkpoint of the model from the [FOMO300K paper](https://arxiv.org/abs/2506.14432).
11
+
12
+ NB: This checkpoint was trained on an earlier version of FOMO300K, which included 11K scans that are no longer part of the dataset.
13
+
14
+ * Network: Resenc Unet B with 102M parameters ([model definition](https://github.com/Sllambias/asparagus/blob/72f333b9b59cf9c2a5689995b72a06d689707fec/asparagus/modules/networks/resenc_unet.py#L119))
15
+ * Pretext Task: [AMAES](https://arxiv.org/abs/2408.00640).
16
+ * Dataset: [FOMO300K](https://huggingface.co/datasets/FOMO-MRI/FOMO300K).
17
+ * Code: [Asparagus](https://github.com/Sllambias/asparagus).
18
+
19
+ ## Finetuning the model for Segmentation tasks
20
+
21
+ To finetune the model with the configs used in the paper, simply use the following command
22
+ ```
23
+ asp_finetune_seg \
24
+ --config-name projects/datapaper/finetune/resenc_[dataset]_fewshot.yaml \
25
+ checkpoint_path=path/to/checkpoint.ckpt
26
+ ```
27
+ where `[dataset]` is one of `{cerebrum, wmh, isles, atlas, sbm}`.
28
+
29
+
30
+ ## Installing Asparagus
31
+
32
+ For instructions on how to install asparagus, preprocess your data, and make your own configs, please refer to the [Asparagus Readme](https://github.com/Sllambias/asparagus).
baselines/resenc_unet_b.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9d698d616d2c97756faa18e2aca5ce823a042c6ff318a9e52a4cc6821ced649
3
+ size 1167178245
baselines/resenc_unet_b_5ch.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f64c0210065f8b96ca0e4433cbd05c69416a683d3b08ea8e7ba2b78b9672ae9
3
+ size 1167207492