jhynsoo/worlabel
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Feat: detection/train 프로젝트 명세 변경에 따른 수정

Browse Source
This commit is contained in:
김진현 2024-09-27 11:18:52 +09:00
parent 2d8609077c
commit 0b38e54b21
6 changed files with 64 additions and 80 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
ai/app/api/yolo/classfication.py
View File

@ -5,7 +5,6 @@ from schemas.predict_response import PredictResponse, LabelData
from schemas.train_report_data import ReportData
from services.load_model import load_classification_model
from services.create_model import save_model
from utils.dataset_utils import split_data
from utils.file_utils import get_dataset_root_path, process_directories_in_cls, process_image_and_label_in_cls, join_path
from utils.slackMessage import send_slack_message
from utils.api_utils import send_data_call_api

116
ai/app/api/yolo/detection.py
View File

@ -1,13 +1,12 @@
from fastapi import APIRouter, HTTPException, Request
from fastapi import APIRouter, HTTPException
from schemas.predict_request import PredictRequest
from schemas.train_request import TrainRequest
from schemas.train_request import TrainRequest, TrainDataInfo
from schemas.predict_response import PredictResponse, LabelData, Shape
from schemas.train_report_data import ReportData
from schemas.train_response import TrainResponse
from services.load_model import load_detection_model
from services.create_model import save_model
from utils.dataset_utils import split_data
from utils.file_utils import get_dataset_root_path, process_directories, process_image_and_label, join_path
from utils.file_utils import get_dataset_root_path, process_directories, join_path, process_image_and_label
from utils.slackMessage import send_slack_message
from utils.api_utils import send_data_call_api
import random
@ -107,71 +106,64 @@ def get_random_color():
async def detection_train(request: TrainRequest):
send_slack_message(f"train 요청{request}", status="success")
# 데이터셋 루트 경로 얻기 (프로젝트 id 기반)
dataset_root_path = get_dataset_root_path(request.project_id)
# 모델 로드
model = get_model(request)
# 이 값을 학습할때 넣으면 이 카테고리들이 학습됨
names = list(request.label_map)
# 데이터 전처리: 학습할 디렉토리 & 데이터셋 를 생성
process_directories(dataset_root_path, names)
# 데이터 전처리: 데이터를 학습데이터와 검증데이터로 분류
train_data, val_data = split_data(request.data, request.ratio)
# 데이터 전처리: 데이터 이미지 및 레이블 다운로드
download_data(train_data, val_data, dataset_root_path, request.label_map)
# 학습
results = run_train(request, model,dataset_root_path)
# best 모델 저장
model_key = save_model(project_id=request.project_id, path=join_path(dataset_root_path, "result", "weights", "best.pt"))
result = results.results_dict
response = TrainResponse(
modelKey=model_key,
precision= result["metrics/precision(B)"],
recall= result["metrics/recall(B)"],
mAP50= result["metrics/mAP50(B)"],
mAP5095= result["metrics/mAP50-95(B)"],
fitness= result["fitness"]
)
send_slack_message(f"train 성공{response}", status="success")
return response
def split_data(data:list[TrainDataInfo], ratio:float):
try:
# 레이블 맵
inverted_label_map = {value: key for key, value in request.label_map.items()} if request.label_map else None
# 데이터셋 루트 경로 얻기
dataset_root_path = get_dataset_root_path(request.project_id)
# 모델 로드
model = get_model(request)
# 학습할 모델 카테고리, 카테고리가 추가되는 경우 추가 작업 필요
model_categories = model.names
# 데이터 전처리
preprocess_dataset(dataset_root_path, model_categories, request.data, request.ratio, inverted_label_map)
# 학습
results = run_train(request, model,dataset_root_path)
# best 모델 저장
model_key = save_model(project_id=request.project_id, path=join_path(dataset_root_path, "result", "weights", "best.pt"))
result = results.results_dict
response = TrainResponse(
modelKey=model_key,
precision= result["metrics/precision(B)"],
recall= result["metrics/recall(B)"],
mAP50= result["metrics/mAP50(B)"],
mAP5095= result["metrics/mAP50-95(B)"],
fitness= result["fitness"]
)
send_slack_message(f"train 성공{response}", status="success")
return response
except HTTPException as e:
raise e
train_size = int(ratio * len(data))
random.shuffle(data)
train_data = data[:train_size]
val_data = data[train_size:]
return train_data, val_data
except Exception as e:
raise HTTPException(status_code=500, detail=str(e))
raise HTTPException(status_code=500, detail="exception in split_data(): " + str(e))
def preprocess_dataset(dataset_root_path, model_categories, data, ratio, label_map):
def download_data(train_data:list[TrainDataInfo], val_data:list[TrainDataInfo], dataset_root_path:str):
try:
# 디렉토리 생성 및 초기화
process_directories(dataset_root_path, model_categories)
# 학습 데이터 분류
train_data, val_data = split_data(data, ratio)
if not train_data or not val_data:
raise HTTPException(status_code=400, detail="data split exception: data size is too small or \"ratio\" has invalid value")
# 학습 데이터 처리
for data in train_data:
process_image_and_label(data, dataset_root_path, "train", label_map)
process_image_and_label(data, dataset_root_path, "train")
# 검증 데이터 처리
for data in val_data:
process_image_and_label(data, dataset_root_path, "val", label_map)
except HTTPException as e:
raise e # HTTP 예외를 다시 발생
process_image_and_label(data, dataset_root_path, "val")
except Exception as e:
raise HTTPException(status_code=500, detail="preprocess dataset exception: " + str(e))
raise HTTPException(status_code=500, detail="exception in download_data(): " + str(e))
def run_train(request, model, dataset_root_path):
try:
@ -201,7 +193,7 @@ def run_train(request, model, dataset_root_path):
# 데이터 전송
send_data_call_api(request.project_id, request.m_id, data)
except Exception as e:
raise HTTPException(status_code=500, detail=f"send_data exception: {e}")
raise HTTPException(status_code=500, detail=f"exception in send_data(): {e}")
# 콜백 등록
model.add_callback("on_train_epoch_start", send_data)
@ -226,6 +218,6 @@ def run_train(request, model, dataset_root_path):
except HTTPException as e:
raise e # HTTP 예외를 다시 발생
except Exception as e:
raise HTTPException(status_code=500, detail=f"run_train exception: {e}")
raise HTTPException(status_code=500, detail=f"exception in run_train(): {e}")

1
ai/app/api/yolo/segmentation.py
View File

@ -6,7 +6,6 @@ from schemas.train_report_data import ReportData
from schemas.train_response import TrainResponse
from services.load_model import load_segmentation_model
from services.create_model import save_model
from utils.dataset_utils import split_data
from utils.file_utils import get_dataset_root_path, process_directories, process_image_and_label, join_path
from utils.slackMessage import send_slack_message
from utils.api_utils import send_data_call_api

2
ai/app/schemas/train_request.py
View File

@ -10,7 +10,7 @@ class TrainRequest(BaseModel):
project_id: int
m_key: str = Field("yolo8", alias="model_key")
m_id: int = Field(..., alias="model_id") # 학습 중 에포크 결과를 보낼때 model_id를 보냄
label_map: dict[int, int] = Field({}, description="모델 레이블 카테고리 idx: 프로젝트 레이블 카테고리 idx , None 일경우 레이블 데이터(프로젝트 레이블)의 idx로 학습")
label_map: dict[str, int] = Field(..., description="프로젝트 레이블 이름: 프로젝트 레이블 pk , None일 경우 모델 레이블 카테고리 idx로 레이블링")
data: List[TrainDataInfo]
ratio: float = 0.8 # 훈련/검증 분할 비율

8
ai/app/utils/dataset_utils.py
View File

@ -1,8 +0,0 @@
import random
def split_data(data:list, ratio:float):
train_size = int(ratio * len(data))
random.shuffle(data)
train_data = data[:train_size]
val_data = data[train_size:]
return train_data, val_data

16
ai/app/utils/file_utils.py
View File

@ -39,7 +39,7 @@ def process_directories(dataset_root_path:str, model_categories:list[str]):
shutil.rmtree(os.path.join(dataset_root_path, "result"))
make_yml(dataset_root_path, model_categories)
def process_image_and_label(data:TrainDataInfo, dataset_root_path:str, child_path:str, label_map:dict[int, int]|None):
def process_image_and_label(data:TrainDataInfo, dataset_root_path:str, child_path:str):
"""이미지 저장 및 레이블 파일 생성"""
# 이미지 url로부터 파일명 분리
img_name = data.image_url.split('/')[-1]
@ -60,11 +60,11 @@ def process_image_and_label(data:TrainDataInfo, dataset_root_path:str, child_pat
# 레이블 -> 학습용 레이블 데이터 파싱 후 생성
if label['task_type'] == "det":
create_detection_train_label(label, label_path, label_map)
create_detection_train_label(label, label_path)
elif label["task_type"] == "seg":
create_segmentation_train_label(label, label_path, label_map)
create_segmentation_train_label(label, label_path)
def create_detection_train_label(label:dict, label_path:str, label_map:dict[int, int]|None):
def create_detection_train_label(label:dict, label_path:str):
with open(label_path, "w") as train_label_txt:
for shape in label["shapes"]:
train_label = []
@ -72,18 +72,18 @@ def create_detection_train_label(label:dict, label_path:str, label_map:dict[int,
y1 = shape["points"][0][1]
x2 = shape["points"][1][0]
y2 = shape["points"][1][1]
train_label.append(str(label_map[shape["group_id"]]) if label_map else str(shape["group_id"])) # label Id
train_label.append(str(shape["group_id"])) # label Id
train_label.append(str((x1 + x2) / 2 / label["imageWidth"])) # 중심 x 좌표
train_label.append(str((y1 + y2) / 2 / label["imageHeight"])) # 중심 y 좌표
train_label.append(str((x2 - x1) / label["imageWidth"])) # 너비
train_label.append(str((y2 - y1) / label["imageHeight"] )) # 높이
train_label_txt.write(" ".join(train_label)+"\n")
def create_segmentation_train_label(label:dict, label_path:str, label_map:dict[int, int]|None):
def create_segmentation_train_label(label:dict, label_path:str):
with open(label_path, "w") as train_label_txt:
for shape in label["shapes"]:
train_label = []
train_label.append(str(label_map[shape["group_id"]]) if label_map else str(shape["group_id"])) # label Id
train_label.append(str(shape["group_id"])) # label Id
for x, y in shape["points"]:
train_label.append(str(x / label["imageWidth"]))
train_label.append(str(y / label["imageHeight"]))
@ -142,3 +142,5 @@ def process_image_and_label_in_cls(data:TrainDataInfo, dataset_root_path:str, ch
# url로부터 이미지 다운로드
urllib.request.urlretrieve(data.image_url, os.path.join(label_path, img_name))
def download_image(url, path):
urllib.request.urlretrieve(url, path)