azaion/ai-training
1
0
Fork 0
mirror of https://github.com/azaion/ai-training.git synced 2026-04-22 03:46:36 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix folder names, refactoring

Browse Source 
move to yolov8m
use checkpoint.txt instead of yaml
This commit is contained in:
Alex Bezdieniezhnykh
2024-06-29 22:18:34 +03:00
parent 3fd72202f2
commit bb1dbfe1e7
10 changed files with 116 additions and 164 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+1 -2
View File
@@ -15,6 +15,5 @@ nvcc --version
python -m pip install --upgrade pip
pip install --upgrade huggingface_hub
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
pip install -q git+https://github.com/THU-MIG/yolov10.git
pip install albumentations
pip install ultralytics albumentations
```
checkpoint.txt
+1
View File
@@ -0,0 +1 @@
2024-06-27 20:51:35
config.py
-22
View File
@@ -1,22 +0,0 @@
import yaml
config_file = 'config.yaml'
class Config:
def __init__(self):
with open(config_file, 'r') as f:
c = yaml.safe_load(f)
self.checkpoint = c['checkpoint']
self.images_dir = c['images_dir']
self.labels_dir = c['labels_dir']
f.close()
def write(self):
with open(config_file, 'w') as f:
d = dict(checkpoint=self.checkpoint,
images_dir=self.images_dir,
labels_dir=self.labels_dir)
yaml.safe_dump(d, f)
f.close()
config.yaml
-3
View File
@@ -1,3 +0,0 @@
checkpoint: 2024-06-18 19:14:02.080664
images_dir: E:\images
labels_dir: E:\labels
constants.py
+18 -4
View File
@@ -1,9 +1,23 @@
import os
from dto.annotationClass import AnnotationClass
current_dataset_dir = os.path.join('datasets', 'zombobase-current')
current_images_dir = os.path.join(current_dataset_dir, 'images')
current_labels_dir = os.path.join(current_dataset_dir, 'labels')
prefix = 'azaion-'
images = 'images'
labels = 'labels'
data_dir = '/azaion/data/raw'
data_images_dir = os.path.join(data_dir, images)
data_labels_dir = os.path.join(data_dir, labels)
processed_dir = '/azaion/data/processed'
processed_images_dir = os.path.join(processed_dir, images)
processed_labels_dir = os.path.join(processed_dir, labels)
datasets_dir = '/azaion/datasets'
models_dir = '/azaion/models'
annotation_classes = AnnotationClass.read_json()
prefix = 'zombobase-'
date_format = '%Y-%m-%d'
checkpoint_file = 'checkpoint.txt'
checkpoint_date_format = '%Y-%m-%d %H:%M:%S'
dataset-visualiser.py
+5 -3
View File
@@ -6,11 +6,13 @@ from dto.imageLabel import ImageLabel
from preprocessing import read_labels
from matplotlib import pyplot as plt
from constants import datasets_dir, prefix
annotation_classes = AnnotationClass.read_json()
dataset_dir = os.path.join('datasets', 'zombobase-2024-06-18', 'train')
images_dir = os.path.join(dataset_dir, 'images')
labels_dir = os.path.join(dataset_dir, 'labels')
cur_dataset = os.path.join(datasets_dir, f'{prefix}2024-06-18', 'train')
images_dir = os.path.join(cur_dataset, 'images')
labels_dir = os.path.join(cur_dataset, 'labels')
for f in os.listdir(images_dir)[35247:]:
image_path = os.path.join(images_dir, f)
preprocessing.py
+26 -67
View File
@@ -1,19 +1,15 @@
import os.path
import time
from datetime import datetime
from datetime import datetime, timedelta
from pathlib import Path
import albumentations as A
import cv2
import numpy as np
from dateutil.relativedelta import relativedelta
from config import Config
from constants import current_images_dir, current_labels_dir, annotation_classes, prefix, date_format, \
current_dataset_dir
from constants import (data_images_dir, data_labels_dir, processed_images_dir, processed_labels_dir,
annotation_classes, checkpoint_file, checkpoint_date_format)
from dto.imageLabel import ImageLabel
config = Config()
def image_processing(img_ann: ImageLabel) -> [ImageLabel]:
transforms = [
@@ -45,12 +41,12 @@ def image_processing(img_ann: ImageLabel) -> [ImageLabel]:
img = ImageLabel(
image=res['image'],
labels=res['bboxes'],
image_path=os.path.join(current_images_dir, f'{name}{path.suffix}'),
labels_path=os.path.join(current_labels_dir, f'{name}.txt')
image_path=os.path.join(processed_images_dir, f'{name}{path.suffix}'),
labels_path=os.path.join(processed_labels_dir, f'{name}.txt')
)
results.append(img)
except Exception as e:
print(f'Error during transformtation: {e}')
print(f'Error during transformation: {e}')
return results
@@ -80,6 +76,7 @@ def read_labels(labels_path) -> [[]]:
str_coordinates = row.split(' ')
class_num = str_coordinates.pop(0)
coordinates = [float(n.replace(',', '.')) for n in str_coordinates]
# noinspection PyTypeChecker
coordinates.append(class_num)
arr.append(coordinates)
return arr
@@ -92,34 +89,24 @@ def process_image(img_ann):
write_result(ImageLabel(
image=img_ann.image,
labels=img_ann.labels,
image_path=os.path.join(current_images_dir, Path(img_ann.image_path).name),
labels_path=os.path.join(current_labels_dir, Path(img_ann.labels_path).name)
image_path=os.path.join(processed_images_dir, Path(img_ann.image_path).name),
labels_path=os.path.join(processed_labels_dir, Path(img_ann.labels_path).name)
))
# os.remove(img_ann.image_path)
# os.remove(img_ann.labels_path)
def get_checkpoint():
if config.checkpoint is not None:
return config.checkpoint
dates = []
for directory in os.listdir('models'):
try:
dates.append(datetime.strptime(directory[len(prefix):], date_format))
except:
continue
if len(dates) == 0:
return datetime.now() - relativedelta(years=1)
else:
return max(dates)
def main():
last_date = checkpoint = get_checkpoint()
checkpoint = datetime.now() - timedelta(days=720)
try:
with open(checkpoint_file, 'r') as f:
checkpoint = datetime.strptime(f.read(), checkpoint_date_format)
except:
pass
last_date = checkpoint
while True:
images = []
with os.scandir(config.images_dir) as imd:
with os.scandir(data_images_dir) as imd:
for image_file in imd:
if not image_file.is_file():
continue
@@ -130,8 +117,8 @@ def main():
for image_file in images:
try:
image_path = os.path.join(config.images_dir, image_file.name)
labels_path = os.path.join(config.labels_dir, f'{Path(image_path).stem}.txt')
image_path = os.path.join(data_images_dir, image_file.name)
labels_path = os.path.join(data_labels_dir, f'{Path(image_path).stem}.txt')
image = cv2.imdecode(np.fromfile(image_path, dtype=np.uint8), cv2.IMREAD_UNCHANGED)
process_image(ImageLabel(
image_path=image_path,
@@ -142,42 +129,14 @@ def main():
except Exception as e:
print(f'Error appeared {e}')
if last_date != checkpoint:
checkpoint = config.checkpoint = last_date
config.write()
checkpoint = last_date
try:
with open(checkpoint_file, 'w') as f:
f.write(datetime.strftime(checkpoint, checkpoint_date_format))
except:
pass
time.sleep(5)
def check_labels():
for label in os.listdir(os.path.join(current_dataset_dir, 'labels')):
with open(os.path.join(current_dataset_dir, 'labels', label), 'r') as f:
lines = f.readlines()
for line in lines:
list_c = line.split(' ')[1:]
for l in list_c:
if float(l) > 1:
print('Error!')
def fix_class(folder):
for label in os.listdir(folder):
if label.startswith('0000'):
with open(os.path.join(folder, label), 'r+') as f:
lines = f.readlines()
truncated = False
for i in range(0, len(lines)):
if len(lines[i]) < 25:
print(lines[i])
truncated = True
lines.pop(i)
if truncated:
f.truncate(0)
f.seek(0)
f.writelines(lines)
f.close()
if __name__ == '__main__':
fix_class('datasets/zombobase-2024-06-18/test/labels')
fix_class('datasets/zombobase-2024-06-18/train/labels')
fix_class('datasets/zombobase-2024-06-18/valid/labels')
# main()
main()
train.py
+25 -23
View File
@@ -3,19 +3,20 @@ from os.path import abspath
import shutil
from datetime import datetime
from pathlib import Path
from ultralytics import YOLOv10
from constants import current_images_dir, current_labels_dir, annotation_classes, prefix, date_format
from ultralytics import YOLO
from constants import processed_images_dir, processed_labels_dir, annotation_classes, prefix, date_format, datasets_dir, models_dir
latest_model = f'models/{prefix}latest.pt'
latest_model = path.join(models_dir, f'{prefix}latest.pt')
today_folder = f'{prefix}{datetime.now():{date_format}}'
today_dataset = path.join(datasets_dir, today_folder)
train_set = 70
valid_set = 20
test_set = 10
def form_dataset():
makedirs(path.join('datasets', today_folder), exist_ok=True)
images = listdir(current_images_dir)
makedirs(today_dataset, exist_ok=True)
images = listdir(processed_images_dir)
train_size = int(len(images) * train_set / 100.0)
valid_size = int(len(images) * valid_set / 100.0)
@@ -28,15 +29,14 @@ def form_dataset():
def move_annotations(images, folder):
today_dataset = path.join('datasets', today_folder)
destination_images = path.join(today_dataset, folder, 'images')
makedirs(destination_images, exist_ok=True)
destination_labels = path.join(today_dataset, folder, 'labels')
makedirs(destination_labels, exist_ok=True)
for image_name in images:
image_path = path.join(current_images_dir, image_name)
image_path = path.join(processed_images_dir, image_name)
label_name = f'{Path(image_name).stem}.txt'
label_path = path.join(current_labels_dir, label_name)
label_path = path.join(processed_labels_dir, label_name)
if not check_label(label_path):
remove(image_path)
else:
@@ -77,38 +77,40 @@ def create_yaml():
lines.append(f'val: valid/images')
lines.append('')
today_yaml = abspath(path.join('datasets', today_folder, 'data.yaml'))
today_yaml = abspath(path.join(today_dataset, 'data.yaml'))
with open(today_yaml, 'w', encoding='utf-8') as f:
f.writelines([f'{line}\n' for line in lines])
def revert_to_current(date):
def revert_to_processed_data(date):
def revert_dir(src_dir, dest_dir):
for file in listdir(src_dir):
s = path.join(src_dir, file)
d = path.join(dest_dir, file)
replace(s, d)
date_dataset = path.join('datasets', f'{prefix}{date}')
current_dataset = path.join('datasets', f'{prefix}current')
date_dataset = path.join(datasets_dir, f'{prefix}{date}')
makedirs(processed_images_dir, exist_ok=True)
makedirs(processed_labels_dir, exist_ok=True)
for subset in ['test', 'train', 'valid']:
revert_dir(path.join(date_dataset, subset, 'images'), path.join(current_dataset, 'images'))
revert_dir(path.join(date_dataset, subset, 'labels'), path.join(current_dataset, 'labels'))
revert_dir(path.join(date_dataset, subset, 'images'), processed_images_dir)
revert_dir(path.join(date_dataset, subset, 'labels'), processed_labels_dir)
shutil.rmtree(date_dataset)
if __name__ == '__main__':
# form_dataset()
# create_yaml()
m = latest_model or 'yolov10x.yaml'
print(f'Initial model: {m}')
model = YOLOv10(latest_model or 'yolov10x.yaml')
model_name = latest_model if path.isfile(latest_model) else 'yolov8m.yaml'
print(f'Initial model: {model_name}')
model = YOLO(model_name)
# cur_folder = path.join(datasets_dir, f'{prefix}2024-06-18')
folder = f'{prefix}2024-06-18'
yaml = abspath(path.join('datasets', folder, 'data.yaml'))
results = model.train(data=yaml, epochs=100, batch=10, imgsz=640, save_period=1)
cur_folder = today_dataset
yaml = abspath(path.join(cur_folder, 'data.yaml'))
results = model.train(data=yaml, epochs=100, batch=55, imgsz=640, save_period=1)
shutil.copy(f'{results.save_dir}/weights/best.pt', latest_model)
shutil.copytree(results.save_dir, f'models/{folder}')
shutil.copytree(results.save_dir, path.join(models_dir, cur_folder))
shutil.rmtree('runs')
shutil.rmtree('models/zombobase-latest')
shutil.rmtree(path.join(models_dir, f'{prefix}latest'))
yolov10x.yaml
-40
View File
@@ -1,40 +0,0 @@
# Parameters
nc: 50 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
# [depth, width, max_channels]
x: [1.00, 1.25, 512]
# YOLOv8.0n backbone
backbone:
# [from, repeats, module, args]
- [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, SCDown, [512, 3, 2]] # 5-P4/16
- [-1, 6, C2fCIB, [512, True]]
- [-1, 1, SCDown, [1024, 3, 2]] # 7-P5/32
- [-1, 3, C2fCIB, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 9
- [-1, 1, PSA, [1024]] # 10
# YOLOv8.0n head
head:
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 6], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2fCIB, [512, True]] # 13
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 4], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 16 (P3/8-small)
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 13], 1, Concat, [1]] # cat head P4
- [-1, 3, C2fCIB, [512, True]] # 19 (P4/16-medium)
- [-1, 1, SCDown, [512, 3, 2]]
- [[-1, 10], 1, Concat, [1]] # cat head P5
- [-1, 3, C2fCIB, [1024, True]] # 22 (P5/32-large)
- [[16, 19, 22], 1, v10Detect, [nc]] # Detect(P3, P4, P5)
yolov8m.yaml
+40
View File
@@ -0,0 +1,40 @@
# Ultralytics YOLO 🚀, GPL-3.0 license
# Parameters
nc: 50 # number of classes
depth_multiple: 0.67 # scales module repeats
width_multiple: 0.75 # scales convolution channels
# YOLOv8.0m backbone
backbone:
# [from, repeats, module, args]
- [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, Conv, [512, 3, 2]] # 5-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, Conv, [768, 3, 2]] # 7-P5/32
- [-1, 3, C2f, [768, True]]
- [-1, 1, SPPF, [768, 5]] # 9
# YOLOv8.0m head
head:
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 6], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2f, [512]] # 13
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 4], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 17 (P3/8-small)
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 12], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [512]] # 20 (P4/16-medium)
- [-1, 1, Conv, [512, 3, 2]]
- [[-1, 9], 1, Concat, [1]] # cat head P5
- [-1, 3, C2f, [768]] # 23 (P5/32-large)
- [[15, 18, 21], 1, Detect, [nc]] # Detect(P3, P4, P5)