notesum.ai

Published at December 9

3D Graph Attention Networks for High Fidelity Pediatric Glioma Segmentation

cs.CV

Released Date: December 9, 2024

Authors: Harish Thangaraj¹, Diya Katariya, Eshaan Joshi, Sangeetha N

Aff.: ¹School of Computer Science and Engineering Vellore Institute of Technology, Chennai

Arxiv: http://arxiv.org/pdf/2412.06743v1

Refer to caption

Parameter	Value
is_debugging	False
all_samples_as_train	False
fold	4
seed	42
max_epochs	75
mednext_size	B
mednext_ksize	3
mednext_ckpt	None
deep_sup	True
batch_size	1
sw_batch_size	2
num_workers	4
roi_x	128
roi_y	128
roi_z	128
infer_overlap	0.5
aug_type	1
loss_type	3
mean_batch	True
learning rate (lr)	$3\times 10^{-4}$
weight decay	$1\times 10^{-6}$
lr_scheduler	cosine-with-warmup
n_gpus	1
pin_memory	True
check_val_every_n_epoch	1
precision	16
amp_backend	native
accumulate_grad_batches	4