Implement Siamese Network for ISIC 2020 Skin Lesion Classification – s4778251 - #284

Open

lgyts wants to merge 38 commits into

shakes76:topic-recognitionfrom

lgyts:topic-recognition

Commits on Oct 29, 2025

update the skeleton
lgyts
committed
aa3e925

Commits on Oct 31, 2025

split the data to balance the ratio
lgyts
committed
da91f04
add Triplet dataset utils
lgyts
committed
bef99a9

Commits on Nov 1, 2025

update the image transform method
lgyts
committed
08473e6
update the loader for test
lgyts
committed
eff8fc0
update params.py which restore all parameters
lgyts
committed
0325de8
use ResNet50 as backbone for feature extraction
lgyts
committed
66426ca
add 4 layer MLP for binary classification, with LeakyReLU activations
lgyts
committed
c6be584
change the hyperparams and create the train_siamese skeleton
lgyts
committed
40365b3
extract features using trained encoder
lgyts
committed
6966a33
train classifier on extracted features
lgyts
committed
857b570
create ignore file for data files, and add main function for train.py
lgyts
committed
dfa4627
fix the path of dataset, change the display of progress check
lgyts
committed
5711b10
ignore the model file and change the datapath
lgyts
committed
a991d76
finish the predict file, load models then predict the class
lgyts
committed
d186476

Commits on Nov 2, 2025

increase the epoch number
lgyts
committed
2ef33a3
restructure the file add the commit and move plot function to utils
lgyts
committed
e55e170
first darft of readme
lgyts
committed
ae5ea4e
upload the images and add the header
lgyts
committed
4b506fb
final update of readme
lgyts
committed
671ffb9

Commits on Nov 3, 2025

Revert "increase the epoch number"
lgyts
committed
540ef33
backroll to increase epoch patch
lgyts
committed
849f8f5
add triplet_val in dataset for train to print after each epoch
lgyts
committed
d52d9e6
add the plot for siamese loss and classify loss
lgyts
committed
de5cb7d
put plot function in utils.py
lgyts
committed
c1713cd
add color jitter for increased robustness
lgyts
committed
92b1b45
add dropout in classifier try to reduce overfitting
lgyts
committed
8824dd2
add L2 normalization and set early stop CLS to prevent overfitting
lgyts
committed
74ef342
decrease dimension from 1000 to 512
lgyts
committed
3d891ec
increase max epoch and add the early stop methods for both
lgyts
committed
3cc79fd
update input, output images, split predict functionality from train.py into predict.py
lgyts
committed
ccd02d8
move extract_features() from predict.py and train.py into utils.py
lgyts
committed
239b92c
centralize and standardize global parameters in params.py, add docstrings across all files
lgyts
committed
f11d27f
first draft of Readme
lgyts
committed
bb3aa9d
add the images for readme
lgyts
committed
342d086
final draft
lgyts
committed
468a688
combine future work and summary in readme
lgyts
committed
2998e92

Commits on Nov 6, 2025

fix typo in readme
lgyts
committed
f8f9326