aidlab/sora_python
7
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

1.新增是否推理获取特征判断

Browse Source 
2.取消搭配不足异常逻辑
This commit is contained in:
zhouchengrong
2024-04-05 17:45:25 +08:00
parent 010d1536eb
commit 726eee86ab
7 changed files with 332 additions and 110 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
app/api/api_outfit_matcher.py
View File

@@ -1,4 +1,5 @@
import logging
import os
import time
from copy import deepcopy
@@ -36,20 +37,22 @@ def outfit_matcher(request_item: OutfitMatcher):
# 查询数据库，分成两批 需要过模型推理的和不需要的
have_features_data = []
no_have_features_data = []
for ai in all_items:
temp_data = deepcopy(all_items)
for td in temp_data:
for sd in search_data:
if ai['item_name'] == sd['item_name']:
ai['features'] = sd['features']
if "features" not in ai.keys():
no_have_features_data.append(ai)
if td['item_name'] == sd['item_name']:
td['features'] = sd['features']
if "features" not in td.keys():
no_have_features_data.append(td)
else:
have_features_data.append(ai)
have_features_data.append(td)
if len(no_have_features_data) > 0:
extracted_features = backbone_service.get_result(no_have_features_data)
# 准备数据
data = deepcopy(all_items) # 做深拷贝 , all_items 是list 可变数组
data = deepcopy(no_have_features_data) # 做深拷贝 , all_items 是list 可变数组
for i, feature in enumerate(extracted_features):
data[i]['features'] = feature
if 'mapped_cate' in data[i].keys():
@@ -68,17 +71,17 @@ def outfit_matcher(request_item: OutfitMatcher):
result = []
start_time = time.time()
for item in request_item['query']:
try:
outfits = fashion_dataset.generate_outfit(item, request_item["topk"], request_item["max_outfits"])
except ValueError as e:
logger.warning(e)
return {"code": 500, "message": f"valueError : {e}", "data": e}
# try:
outfits = fashion_dataset.generate_outfit(item, request_item["topk"], request_item["max_outfits"])
# except ValueError as e:
# logger.warning(e)
# return {"code": 500, "message": f"valueError : {e}", "data": e}
scores = service.get_result(outfits, prepared_feature)
if request_item['is_best']:
best_outfits, best_scores = service.visualize(outfits, scores, request_item["topk"], best=True,
# output_path=os.path.join(r"E:\workspace\outfit_matcher\2024 SS Outfit", f"{item['item_name']}_best_{param['topk']}.png")
# output_path=rf"E:\workspace\trinity_client_mixi\app\service\outfit_matcher\output_outfit\{item['item_name']}_best_{request_item['topk']}.png"
)
result.append({"outfits": best_outfits, "scores": best_scores})
else:

3
app/core/config.py
View File

@@ -35,7 +35,8 @@ ATT_TRITON_PORT = "10020"
MILVUS_URL = "http://10.1.1.240:19530"
DEBUG = 2
DEBUG = 1
SHOW_OR_SAVE_result_image = False
# service env : 1
# pycharm debug : 2

8
app/service/outfit_matcher/dataset.py
View File

@@ -86,8 +86,8 @@ class FashionDataset(object):
outerwear = random.choice(self.cate2item['outerwear'])
outfit.append(outerwear)
outfit_list.append(tuple(outfit))
if len(outfit_list) < topk:
raise ValueError(f"Cannot generate more than {topk} outfits!")
# if len(outfit_list) < topk:
# raise ValueError(f"Cannot generate more than {topk} outfits!")
return outfit_list
elif given_cate == 'outerwear':
@@ -112,8 +112,8 @@ class FashionDataset(object):
outfit = [query_item] + list(top_bottom)
outfit_list.append(tuple(outfit))
if len(outfit_list) < topk:
raise ValueError(f"Cannot generate more than {topk} outfits!")
# if len(outfit_list) < topk:
# raise ValueError(f"Cannot generate more than {topk} outfits!")
return outfit_list
elif given_cate == 'all-body':

77
app/service/outfit_matcher/outfit_evaluator.py
View File

@@ -197,45 +197,46 @@ class OutfitMatcher(object):
outfits = [outfits[i] for i in sorted_indices] # 最好或最差的五个
scores = scores[sorted_indices] # 这五个的分数
return outfits, scores.tolist()
if SHOW_OR_SAVE_result_image:
# 设置子图的行列数
num_rows = len(outfits)
num_cols = max([len(x) for x in outfits]) + 1 # 一个是图片，一个是分数
# # 设置子图的行列数
# num_rows = len(outfits)
# num_cols = max([len(x) for x in outfits]) + 1 # 一个是图片，一个是分数
#
# # 创建一个新的图像，并指定子图的行列数
# fig, axes = plt.subplots(num_rows, num_cols, figsize=(8, 15))
#
# title = f"Best {topk} Outfits" if best else f"Worst {topk} Outfits"
# fig.suptitle(title, fontsize=16)
#
# # 遍历每套outfit并将其显示在对应的子图中
# for i, (outfit, score) in enumerate(zip(outfits, scores)):
# # 显示分数
# axes[i, 0].text(0.1, 0.5, f"Score: {score:.4f}", fontsize=12)
# axes[i, 0].axis("off")
# # 显示图片
# for j, item in enumerate(outfit):
# img = self.load_image(item['image_path']) # 读取图片
# axes[i, j + 1].imshow(img) # 在对应的子图中显示图片
# axes[i, j + 1].axis('off') # 关闭坐标轴
# axes[i, j + 1].set_title(item["semantic_category"], fontsize=10)
# for j in range(len(outfit), num_cols):
# axes[i, j].axis("off")
#
# # 在每一行的底部添加一条横线
# axes[i, 0].axhline(y=0, color='black', linewidth=1)
# # 隐藏最后一行的横线
# axes[-1, 0].axhline(y=0, color='white', linewidth=1)
#
# # 调整布局
# plt.subplots_adjust(wspace=0.1, hspace=0.1)
# plt.tight_layout()
#
# if output_path:
# plt.savefig(output_path)
# else:
# plt.show()
# 创建一个新的图像，并指定子图的行列数
fig, axes = plt.subplots(num_rows, num_cols, figsize=(8, 15))
title = f"Best {topk} Outfits" if best else f"Worst {topk} Outfits"
fig.suptitle(title, fontsize=16)
# 遍历每套outfit并将其显示在对应的子图中
for i, (outfit, score) in enumerate(zip(outfits, scores)):
# 显示分数
axes[i, 0].text(0.1, 0.5, f"Score: {score:.4f}", fontsize=12)
axes[i, 0].axis("off")
# 显示图片
for j, item in enumerate(outfit):
img = self.load_image(item['image_path']) # 读取图片
axes[i, j + 1].imshow(img) # 在对应的子图中显示图片
axes[i, j + 1].axis('off') # 关闭坐标轴
axes[i, j + 1].set_title(item["semantic_category"], fontsize=10)
for j in range(len(outfit), num_cols):
axes[i, j].axis("off")
# 在每一行的底部添加一条横线
axes[i, 0].axhline(y=0, color='black', linewidth=1)
# 隐藏最后一行的横线
axes[-1, 0].axhline(y=0, color='white', linewidth=1)
# 调整布局
plt.subplots_adjust(wspace=0.1, hspace=0.1)
plt.tight_layout()
if output_path:
plt.savefig(output_path)
else:
plt.show()
return outfits, scores.tolist()
class OutfitMatcherHon(OutfitMatcher):

113
app/service/outfit_matcher/service.py
View File

@@ -1,57 +1,84 @@
import json
import os
from pprint import pprint
import numpy as np
import time
from copy import deepcopy
from pymilvus import MilvusClient
from app.core.config import *
from app.service.outfit_matcher.dataset import FashionDataset
from app.service.outfit_matcher.outfit_evaluator import OutfitMaterTypeAware, Backbone
logger = logging.getLogger()
if __name__ == '__main__':
with open("./test_param/recommendation_test.json", "r") as f:
param = json.load(f)
fashion_dataset = FashionDataset(param["database"])
with open("./test_param/test.json", "r") as f:
request_item = json.load(f)
request_item = dict(request_item)
for i in range(len(request_item['query'])):
request_item['query'][i] = dict(request_item['query'][i])
for i in range(len(request_item['database'])):
request_item['database'][i] = dict(request_item['database'][i])
fashion_dataset = FashionDataset(request_item['database'])
backbone_service = Backbone()
service = OutfitMaterTypeAware()
# read feature from vector database
all_items = param["query"] + param["database"]
unextracted_item = []
all_items = request_item["query"] + request_item["database"]
prepared_feature = {}
# 拿到所有需要提取特征的图片
for item in all_items:
if f'{item["item_name"]}.npy' not in os.listdir("feature"):
unextracted_item.append(item)
if len(unextracted_item) > 0:
# 通过backbone模型提取图片特征
extracted_features = backbone_service.get_result(unextracted_item)
for i, item in enumerate(unextracted_item):
# save features
# 链接milvus
# TODO
np.save(f'feature/{item["item_name"]}.npy', extracted_features[i])
# 存入数据库
# 关闭链接
# 连接milvus
client = MilvusClient(uri=MILVUS_URL, token="root:Milvus", db_name="mixi")
search_data = client.get(collection_name='mixi_outfit', ids=[item['item_name'] for item in all_items])
# TODO 读取本次任务需要的图片特征
for item in all_items:
if item["item_name"] not in prepared_feature.keys():
prepared_feature[item["item_name"]] = np.load(f'feature/{item["item_name"]}.npy')
# 查询数据库，分成两批 需要过模型推理的和不需要的
have_features_data = []
no_have_features_data = []
temp_data = deepcopy(all_items)
for td in temp_data:
for sd in search_data:
if td['item_name'] == sd['item_name']:
td['features'] = sd['features']
if "features" not in td.keys():
no_have_features_data.append(td)
else:
have_features_data.append(td)
if len(no_have_features_data) > 0:
extracted_features = backbone_service.get_result(no_have_features_data)
# 准备数据
data = deepcopy(no_have_features_data) # 做深拷贝 , all_items 是list 可变数组
for i, feature in enumerate(extracted_features):
data[i]['features'] = feature
if 'mapped_cate' in data[i].keys():
del data[i]['mapped_cate']
# 存入数据
res = client.insert(collection_name="mixi_outfit", data=data)
for d in data:
prepared_feature[d['item_name']] = d['features']
for hfd in have_features_data:
prepared_feature[hfd['item_name']] = hfd['features']
# 断开连接
client.close()
result = []
start_time = time.time()
for item in request_item['query']:
# try:
outfits = fashion_dataset.generate_outfit(item, request_item["topk"], request_item["max_outfits"])
# except ValueError as e:
# logger.warning(e)
# return {"code": 500, "message": f"valueError : {e}", "data": e}
# 开始服装搭配任务
for item in param["query"]:
# 根据一定规则生成outfit
outfits = fashion_dataset.generate_outfit(item, param["topk"], param["max_outfits"])
# 根据模型对生成的outfit打分
scores = service.get_result(outfits, prepared_feature)
# 对评分排序，拿到最好的topk个outfit输出
sorted_indices = np.argsort(scores)[:param["topk"]] # type-aware
best_outfits = [outfits[i] for i in sorted_indices] # 最好的五个
# 结果可视化
# service.visualize(outfits, scores, param["topk"], best=True,
# output_path=os.path.join(r"D:\PhD_Study\MIXI\mitu\image\123",
# f"{item['item_name']}_best_{param['topk']}.png"))
# service.visualize(outfits, scores, param["topk"], best=False,
# output_path=os.path.join(r"D:\PhD_Study\MIXI\mitu\image\123",
# f"{item['item_name']}_worst_{param['topk']}.png"))
if request_item['is_best']:
best_outfits, best_scores = service.visualize(outfits, scores, request_item["topk"], best=True,
output_path=rf"E:\workspace\trinity_client_mixi\app\service\outfit_matcher\output_outfit\{item['item_name']}_best_{request_item['topk']}.png"
)
result.append({"outfits": best_outfits, "scores": best_scores})
else:
bad_outfits, bad_scores = service.visualize(outfits, scores, request_item["topk"], best=False,
# output_path=os.path.join(r"E:\workspace\outfit_matcher\2024 SS Outfit", f"{item['item_name']}_worst_{param['topk']}.png")
)
result.append({"outfits": bad_outfits, "scores": bad_scores})
logger.info(f"run time is : {time.time() - start_time}")

2
app/service/outfit_matcher/test_param/recommendation_test.json
View File

@@ -1,6 +1,6 @@
{
"topk": 5,
"max_outfits": 200,
"max_outfits": 10,
"is_best": true,
"query": [
{

210
app/service/outfit_matcher/test_param/test.json
View File

@@ -1,19 +1,209 @@
{
"topk": 1,
"max_outfits": 5,
"is_best": true,
"query": [
"topk": 5,
"max_outfits": 10,
"database": [
{
"image_path": "mi-tu/26/BOTTOM/PANTS/MKTS27000_0BLK.jpg/3f4676db-98a1-44d4-947f-9d1f59828629.jpg",
"image_path": "mi-tu/26/BOTTOM/PANTS/MKTS27000_0BLK.jpg/MKTS27000_0BLK.jpg",
"item_name": "MKTS27000",
"semantic_category": "BOTTOM/PANTS"
},
{
"image_path": "mi-tu/26/TOP/BLOUSE/MKTS27002_0WHT.jpg/MKTS27002_0WHT.jpg",
"item_name": "MKTS27002",
"semantic_category": "TOP/BLOUSE"
},
{
"image_path": "mi-tu/26/BOTTOM/SHORTS/MKTS27001_0BLK.jpg/MKTS27001_0BLK.jpg",
"item_name": "MKTS27001",
"semantic_category": "BOTTOM/SHORTS"
},
{
"image_path": "mi-tu/26/BOTTOM/PANTS/MKTS27004_0BGE.jpg/MKTS27004_0BGE.jpg",
"item_name": "MKTS27004",
"semantic_category": "BOTTOM/PANTS"
},
{
"image_path": "mi-tu/26/TOP/BLOUSE/MKTS27002_0BLK.jpg/MKTS27002_0BLK.jpg",
"item_name": "MKTS27002",
"semantic_category": "TOP/BLOUSE"
},
{
"image_path": "mi-tu/26/OUTERWEAR/BLAZER/MKTS27008_0BLK.jpg/MKTS27008_0BLK.jpg",
"item_name": "MKTS27008",
"semantic_category": "OUTERWEAR/BLAZER"
},
{
"image_path": "mi-tu/26/BOTTOM/PANTS/MKTS27009_0BLK.jpg/MKTS27009_0BLK.jpg",
"item_name": "MKTS27009",
"semantic_category": "BOTTOM/PANTS"
},
{
"image_path": "mi-tu/26/BOTTOM/PANTS/MKTS27004_0BLU.jpg/MKTS27004_0BLU.jpg",
"item_name": "MKTS27004",
"semantic_category": "BOTTOM/PANTS"
},
{
"image_path": "mi-tu/26/TOP/VEST/MKTS27011_0BLK.jpg/MKTS27011_0BLK.jpg",
"item_name": "MKTS27011",
"semantic_category": "TOP/VEST"
},
{
"image_path": "mi-tu/26/BOTTOM/PANTS/MKTS27009_0FUS.jpg/MKTS27009_0FUS.jpg",
"item_name": "MKTS27009",
"semantic_category": "BOTTOM/PANTS"
},
{
"image_path": "mi-tu/26/OUTERWEAR/GILET/MKTS27003_0BGE.jpg/MKTS27003_0BGE.jpg",
"item_name": "MKTS27003",
"semantic_category": "OUTERWEAR/GILET"
},
{
"image_path": "mi-tu/26/OUTERWEAR/BLAZER/MKTS27010_0WHT.jpg/MKTS27010_0WHT.jpg",
"item_name": "MKTS27010",
"semantic_category": "OUTERWEAR/BLAZER"
},
{
"image_path": "mi-tu/26/TOP/VEST/MKTS27011_0CMY.jpg/MKTS27011_0CMY.jpg",
"item_name": "MKTS27011",
"semantic_category": "TOP/VEST"
},
{
"image_path": "mi-tu/26/BOTTOM/SHORTS/MKTS27013_0ORG.jpg/MKTS27013_0ORG.jpg",
"item_name": "MKTS27013",
"semantic_category": "BOTTOM/SHORTS"
},
{
"image_path": "mi-tu/26/OUTERWEAR/GILET/MKTS27015_0BLK.jpg/MKTS27015_0BLK.jpg",
"item_name": "MKTS27015",
"semantic_category": "OUTERWEAR/GILET"
},
{
"image_path": "mi-tu/26/OUTERWEAR/GILET/MKTS27015_0CMY.jpg/MKTS27015_0CMY.jpg",
"item_name": "MKTS27015",
"semantic_category": "OUTERWEAR/GILET"
},
{
"image_path": "mi-tu/26/OUTERWEAR/BLAZER/MKTS27010_0DBL.jpg/MKTS27010_0DBL.jpg",
"item_name": "MKTS27010",
"semantic_category": "OUTERWEAR/BLAZER"
},
{
"image_path": "mi-tu/26/BOTTOM/SHORTS/MKTS27013_0BLU.jpg/MKTS27013_0BLU.jpg",
"item_name": "MKTS27013",
"semantic_category": "BOTTOM/SHORTS"
},
{
"image_path": "mi-tu/26/OUTERWEAR/JACKET/MKTS27012_0ORG.jpg/MKTS27012_0ORG.jpg",
"item_name": "MKTS27012",
"semantic_category": "OUTERWEAR/JACKET"
},
{
"image_path": "mi-tu/26/BOTTOM/SHORTS/MKTS27016_0CMY.jpg/MKTS27016_0CMY.jpg",
"item_name": "MKTS27016",
"semantic_category": "BOTTOM/SHORTS"
},
{
"image_path": "mi-tu/26/BOTTOM/SHORTS/MKTS27016_0BLK.jpg/MKTS27016_0BLK.jpg",
"item_name": "MKTS27016",
"semantic_category": "BOTTOM/SHORTS"
},
{
"image_path": "mi-tu/26/OUTERWEAR/WINDBREAKER/MKTS27017_0GRN.jpg/MKTS27017_0GRN.jpg",
"item_name": "MKTS27017",
"semantic_category": "OUTERWEAR/WINDBREAKER"
},
{
"image_path": "mi-tu/26/OUTERWEAR/JACKET/MKTS27012_0BLU.jpg/MKTS27012_0BLU.jpg",
"item_name": "MKTS27012",
"semantic_category": "OUTERWEAR/JACKET"
},
{
"image_path": "mi-tu/26/TOP/SHIRT/MKTS27018_0WHT.jpg/MKTS27018_0WHT.jpg",
"item_name": "MKTS27018",
"semantic_category": "TOP/SHIRT"
},
{
"image_path": "mi-tu/26/TOP/SHIRT/MKTS27018_0BLK.jpg/MKTS27018_0BLK.jpg",
"item_name": "MKTS27018",
"semantic_category": "TOP/SHIRT"
},
{
"image_path": "mi-tu/26/OUTERWEAR/BLAZER/MKTS27019_0BLK.jpg/MKTS27019_0BLK.jpg",
"item_name": "MKTS27019",
"semantic_category": "OUTERWEAR/BLAZER"
},
{
"image_path": "mi-tu/26/OUTERWEAR/WINDBREAKER/MKTS27017_0PNK.jpg/MKTS27017_0PNK.jpg",
"item_name": "MKTS27017",
"semantic_category": "OUTERWEAR/WINDBREAKER"
},
{
"image_path": "mi-tu/26/OUTERWEAR/BLAZER/MKTS27019_0PNK.jpg/MKTS27019_0PNK.jpg",
"item_name": "MKTS27019",
"semantic_category": "OUTERWEAR/BLAZER"
},
{
"image_path": "mi-tu/26/BOTTOM/PANTS/MKTS27027_0DBL.jpg/MKTS27027_0DBL.jpg",
"item_name": "MKTS27027",
"semantic_category": "BOTTOM/PANTS"
},
{
"image_path": "mi-tu/26/BOTTOM/PANTS/MKTS27027_0PNK.jpg/MKTS27027_0PNK.jpg",
"item_name": "MKTS27027",
"semantic_category": "BOTTOM/PANTS"
}
],
"database": [
"query": [
{
"image_path": "mi-tu/26/TOP/BLOUSE/MKTS27002_0WHT.jpg/131cc29e-8f70-4134-a0e8-82f826b00058.jpg",
"item_name": "MKTS27002",
"semantic_category": "TOP/BLOUSE"
"image_path": "mi-tu/26/BOTTOM/SHORTS/MKTS27016_0BLK.jpg/MKTS27016_0BLK.jpg",
"item_name": "MKTS27016",
"semantic_category": "BOTTOM/SHORTS"
},
{
"image_path": "mi-tu/26/OUTERWEAR/WINDBREAKER/MKTS27017_0GRN.jpg/MKTS27017_0GRN.jpg",
"item_name": "MKTS27017",
"semantic_category": "OUTERWEAR/WINDBREAKER"
},
{
"image_path": "mi-tu/26/OUTERWEAR/JACKET/MKTS27012_0BLU.jpg/MKTS27012_0BLU.jpg",
"item_name": "MKTS27012",
"semantic_category": "OUTERWEAR/JACKET"
},
{
"image_path": "mi-tu/26/TOP/SHIRT/MKTS27018_0WHT.jpg/MKTS27018_0WHT.jpg",
"item_name": "MKTS27018",
"semantic_category": "TOP/SHIRT"
},
{
"image_path": "mi-tu/26/TOP/SHIRT/MKTS27018_0BLK.jpg/MKTS27018_0BLK.jpg",
"item_name": "MKTS27018",
"semantic_category": "TOP/SHIRT"
},
{
"image_path": "mi-tu/26/OUTERWEAR/BLAZER/MKTS27019_0BLK.jpg/MKTS27019_0BLK.jpg",
"item_name": "MKTS27019",
"semantic_category": "OUTERWEAR/BLAZER"
},
{
"image_path": "mi-tu/26/OUTERWEAR/WINDBREAKER/MKTS27017_0PNK.jpg/MKTS27017_0PNK.jpg",
"item_name": "MKTS27017",
"semantic_category": "OUTERWEAR/WINDBREAKER"
},
{
"image_path": "mi-tu/26/OUTERWEAR/BLAZER/MKTS27019_0PNK.jpg/MKTS27019_0PNK.jpg",
"item_name": "MKTS27019",
"semantic_category": "OUTERWEAR/BLAZER"
},
{
"image_path": "mi-tu/26/BOTTOM/PANTS/MKTS27027_0DBL.jpg/MKTS27027_0DBL.jpg",
"item_name": "MKTS27027",
"semantic_category": "BOTTOM/PANTS"
},
{
"image_path": "mi-tu/26/BOTTOM/PANTS/MKTS27027_0PNK.jpg/MKTS27027_0PNK.jpg",
"item_name": "MKTS27027",
"semantic_category": "BOTTOM/PANTS"
}
]
],
"is_best": true
}