This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
def generateImagesOfVariousSizes(numberOfPoints): | |
directory = 'data/countingVariousSizes/' + str(numberOfPoints) + '/' | |
os.makedirs(directory, exist_ok=True) | |
#Create 5,000 images of this class | |
for j in tnrange(5000): | |
path = directory + str(j) + '.png' | |
#Get points | |
x, y = createNonOverlappingPoints(numberOfPoints) | |
#Create plot |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
learner = create_cnn(data, models.resnet34, metrics=error_rate) | |
learner.fit_one_cycle(15, max_lr=slice(1e-4, 1e-2)) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
learner = create_cnn(data, models.resnet34, metrics=error_rate) | |
learner.fit_one_cycle(3) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
path = 'data/counting' | |
np.random.seed(42) | |
data = ImageDataBunch.from_folder(path, train=".", valid_pct=0.2, | |
ds_tfms=get_transforms(), size=224, num_workers=4).normalize(imagenet_stats) | |
data.show_batch(rows=3, figsize=(7,8)) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
learner = create_cnn(data, models.resnet34, metrics=error_rate) | |
learner.fit_one_cycle(3) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
path = 'data/counting' | |
data = ImageDataBunch.from_folder(path, train=".", valid_pct=0.2, | |
ds_tfms=get_transforms(), size=224, num_workers=4).normalize(imagenet_stats) | |
data.show_batch(rows=3, figsize=(7,8)) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import random | |
import numpy as np | |
import matplotlib.pyplot as plt | |
def createNonOverlappingPoints(numElements): | |
x = np.zeros((numElements)) + 2 #Place the cirlces offscreen | |
y = np.zeros((numElements)) + 2 #Place the circles offscreen | |
for i in range(0, numElements): | |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
learner.fit_one_cycle(15, max_lr=slice(1e-4)) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
learner.save('stage-1') | |
learner.unfreeze() | |
learner.lr_find() | |
learner.recorder.plot() |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
learner = create_cnn(data, models.resnet34, metrics=error_rate) | |
learner.fit_one_cycle(5) |