Getting the Data

from fastai.vision import *

path = untar_data('https://s3.amazonaws.com/fast-ai-imageclas/imagenette2') 

lbl_dict = dict(
    n01440764='tench',
    n02102040='English springer',
    n02979186='cassette player',
    n03000684='chain saw',
    n03028079='church',
    n03394916='French horn',
    n03417042='garbage truck',
    n03425413='gas pump',
    n03445777='golf ball',
    n03888257='parachute'
)
def lbl_trfm(o): 
  return lbl_dict[str(o).split('/')[-2]]

tfms = get_transforms()

data = ImageList.from_folder(path).split_by_folder(valid='val').label_from_func(lbl_trfm).transform(tfms, size=320).databunch(bs=32).normalize(imagenet_stats)

data

ImageDataBunch;

Train: LabelList (9469 items)
x: ImageList
Image (3, 320, 320),Image (3, 320, 320),Image (3, 320, 320),Image (3, 320, 320),Image (3, 320, 320)
y: CategoryList
parachute,parachute,parachute,parachute,parachute
Path: /root/.fastai/data/imagenette2;

Valid: LabelList (3925 items)
x: ImageList
Image (3, 320, 320),Image (3, 320, 320),Image (3, 320, 320),Image (3, 320, 320),Image (3, 320, 320)
y: CategoryList
parachute,parachute,parachute,parachute,parachute
Path: /root/.fastai/data/imagenette2;

Test: None

data.show_batch(5, figsize=(16,16))

Training Model

if torch.cuda.is_available():
  torch.cuda.init()

from fastai.callbacks import *

learn = cnn_learner(data, models.resnet50, metrics=[accuracy], callback_fns=[CSVLogger,ShowGraph])

learn.lr_find()

LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.

learn.recorder.plot()

learn.freeze()
learn.fit_one_cycle(2,max_lr=1e-2)

epoch	train_loss	valid_loss	accuracy	time
0	0.578849	0.410253	0.879490	06:13
1	0.187801	0.092428	0.968662	06:03

learn.unfreeze()

learn.fit_one_cycle(3,max_lr=slice(1e-4,1e-3,1e-2))

epoch	train_loss	valid_loss	accuracy	time
0	0.471138	0.314138	0.898344	07:54
1	0.288256	0.203786	0.935287	07:50
2	0.135715	0.108479	0.963822	07:46

Results

interp = ClassificationInterpretation.from_learner(learn)
losses,idxs = interp.top_losses()

interp.plot_confusion_matrix(figsize=(8,8))

interp.plot_top_losses(9, figsize=(16,16))

Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).