issues
search
truc-h-nguyen
/
Toddler-activity-suggestions
1
stars
0
forks
source link
[dev] add train loop with hyper parameters to change
#13
Open
nickvazz
opened
2 years ago
nickvazz
commented
2 years ago
TODO:
[ ] wrap preprocessing (classes and the image size) in a function
[ ] wrap augmentation setup and application to data in function
[ ] wrap our model generation (with the extra params changed) in a function
[ ] wrap our training in function
[ ] wrap our evaluation metrics (IoU etc) in functions
hyper parameters and metadata to track
[ ] optimizer ("rmsprop")
[ ] loss ('binary_crossentropy')
[ ] batch size
[ ] image size
[ ]
filters
[ ]
num_classes
[ ]
epochs
[ ] turn
residual
on / off
[ ] activation functions
[ ] augmentations to use and their values
[ ] which classes we used
[ ] how many samples in each class
TODO:
hyper parameters and metadata to track
filters
num_classes
epochs
residual
on / off