搭配服务修改

app/api/api_attribute.py

@@ -15,5 +15,4 @@ router = APIRouter()
 def attribute(request_data: AttributeModel):
     service = AttributeRecognition()
     response = service.attribute(const, request_data)
-    logger.info("test")
     return {"code": 200, "message": "ok", "data": response}

app/api/api_outfit_matcher.py

@@ -1,10 +1,13 @@
 import logging
 import time
+from copy import deepcopy
 
 from fastapi import APIRouter
+from pymilvus import MilvusClient
+
 from app.schemas.outfit_matcher import OutfitMatcher
 from app.service.outfit_matcher.dataset import FashionDataset
-from app.service.outfit_matcher.outfit_evaluator import OutfitMaterTypeAware
+from app.service.outfit_matcher.outfit_evaluator import OutfitMaterTypeAware, Backbone
 from app.service.utils.decorator import RunTime
 
 logger = logging.getLogger()
@@ -19,16 +22,28 @@ def outfit_matcher(request_item: OutfitMatcher):
         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])
-
-    # try:
     fashion_dataset = FashionDataset(request_item['database'])
+    backbone_service = Backbone()
     service = OutfitMaterTypeAware()
+    all_items = request_item["query"] + request_item["database"]
+    prepared_feature = {}
+    extracted_features = backbone_service.get_result(all_items)
+    data = deepcopy(all_items)  # 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']
+
+    client = MilvusClient(uri="http://10.1.1.240:19530", token="root:Milvus", db_name="mixi")
+    res = client.insert(collection_name="mixi_outfit", data=data)
+    client.close()
+    for d in data:
+        prepared_feature[d['item_name']] = d['features']
     result = []
     start_time = time.time()
     for item in request_item['query']:
         outfits = fashion_dataset.generate_outfit(item, request_item["topk"], request_item["max_outfits"])
-        scores, features = service.get_result(outfits)
+        scores = service.get_result(outfits, prepared_feature)
-        # save features in databases
 
         if request_item['is_best']:
             best_outfits, best_scores = service.visualize(outfits, scores, request_item["topk"], best=True,

app/core/config.py

@@ -33,17 +33,19 @@ OM_TRITON_PORT = "10010"
 ATT_TRITON_IP = "10.1.1.240"
 ATT_TRITON_PORT = "10020"
 
-# service env
+DEBUG = 1
-LOGS_PATH = "app/logs/errors.log"
+# service env  : 1
-FASHION_CATEGORIES = "app/service/outfit_matcher/config/fashion_categories.json"
+# pycharm debug  : 2
-FASHION_CATEGORIES_MAPPING = "app/service/outfit_matcher/config/fashion_category_mapping.json"
 
-# pycharm debug
+if DEBUG == 1:
-# LOGS_PATH = "logs/errors.log"
+    LOGS_PATH = "app/logs/errors.log"
-# FASHION_CATEGORIES = "service/outfit_matcher/config/fashion_categories.json"
+    FASHION_CATEGORIES = "app/service/outfit_matcher/config/fashion_categories.json"
-# FASHION_CATEGORIES_MAPPING = "service/outfit_matcher/config/fashion_category_mapping.json"
+    FASHION_CATEGORIES_MAPPING = "app/service/outfit_matcher/config/fashion_category_mapping.json"
-
+elif DEBUG == 2:
-
+    LOGS_PATH = "logs/errors.log"
-# LOGS_PATH = "app/logs/errors.log"
+    FASHION_CATEGORIES = "service/outfit_matcher/config/fashion_categories.json"
-# FASHION_CATEGORIES = "./config/fashion_categories.json"
+    FASHION_CATEGORIES_MAPPING = "service/outfit_matcher/config/fashion_category_mapping.json"
-# FASHION_CATEGORIES_MAPPING = "./config/fashion_category_mapping.json"
+elif DEBUG == 3:
+    LOGS_PATH = "app/logs/errors.log"
+    FASHION_CATEGORIES = "./config/fashion_categories.json"
+    FASHION_CATEGORIES_MAPPING = "./config/fashion_category_mapping.json"

app/service/outfit_matcher/outfit_evaluator.py

@@ -197,43 +197,45 @@ 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
 
-        # 设置子图的行列数
+        return outfits, scores.tolist()
-        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
+        # # 设置子图的行列数
-        fig, axes = plt.subplots(num_rows, num_cols, figsize=(8, 15))
+        # num_rows = len(outfits)
-
+        # num_cols = max([len(x) for x in outfits]) + 1  # 一个是图片，一个是分数
-        title = f"Best {topk} Outfits" if best else f"Worst {topk} Outfits"
+        #
-        fig.suptitle(title, fontsize=16)
+        # # 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))
-        # 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)):
+        # title = f"Best {topk} Outfits" if best else f"Worst {topk} Outfits"
-            # 显示分数
+        # fig.suptitle(title, fontsize=16)
-            axes[i, 0].text(0.1, 0.5, f"Score: {score:.4f}", fontsize=12)
+        #
-            axes[i, 0].axis("off")
+        # # 遍历每套outfit并将其显示在对应的子图中
-            # 显示图片
+        # for i, (outfit, score) in enumerate(zip(outfits, scores)):
-            for j, item in enumerate(outfit):
+        #     # 显示分数
-                img = self.load_image(item['image_path'])  # 读取图片
+        #     axes[i, 0].text(0.1, 0.5, f"Score: {score:.4f}", fontsize=12)
-                axes[i, j + 1].imshow(img)  # 在对应的子图中显示图片
+        #     axes[i, 0].axis("off")
-                axes[i, j + 1].axis('off')  # 关闭坐标轴
+        #     # 显示图片
-                axes[i, j + 1].set_title(item["semantic_category"], fontsize=10)
+        #     for j, item in enumerate(outfit):
-            for j in range(len(outfit), num_cols):
+        #         img = self.load_image(item['image_path'])  # 读取图片
-                axes[i, j].axis("off")
+        #         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
-            # 在每一行的底部添加一条横线
+        #         axes[i, j + 1].set_title(item["semantic_category"], fontsize=10)
-            axes[i, 0].axhline(y=0, color='black', linewidth=1)
+        #     for j in range(len(outfit), num_cols):
-        # 隐藏最后一行的横线
+        #         axes[i, j].axis("off")
-        axes[-1, 0].axhline(y=0, color='white', linewidth=1)
+        #
-
+        #     # 在每一行的底部添加一条横线
-        # 调整布局
+        #     axes[i, 0].axhline(y=0, color='black', linewidth=1)
-        plt.subplots_adjust(wspace=0.1, hspace=0.1)
+        # # 隐藏最后一行的横线
-        plt.tight_layout()
+        # axes[-1, 0].axhline(y=0, color='white', linewidth=1)
-
+        #
-        if output_path:
+        # # 调整布局
-            plt.savefig(output_path)
+        # plt.subplots_adjust(wspace=0.1, hspace=0.1)
-        else:
+        # plt.tight_layout()
-            plt.show()
+        #
+        # if output_path:
+        #     plt.savefig(output_path)
+        # else:
+        #     plt.show()
 
 
 class OutfitMatcherHon(OutfitMatcher):

app/service/outfit_matcher/service.py

@@ -1,37 +1,57 @@
 import json
 import os
 from pprint import pprint
+import numpy as np
 
 from app.service.outfit_matcher.dataset import FashionDataset
-from app.service.outfit_matcher.outfit_evaluator import OutfitMaterTypeAware
+from app.service.outfit_matcher.outfit_evaluator import OutfitMaterTypeAware, Backbone
 
 if __name__ == '__main__':
     with open("./test_param/recommendation_test.json", "r") as f:
         param = json.load(f)
     fashion_dataset = FashionDataset(param["database"])
+    backbone_service = Backbone()
     service = OutfitMaterTypeAware()
-    best_list = []
+
-    bad_list = []
+    # read feature from vector database
-    for item in param["query"]:
+    all_items = param["query"] + param["database"]
-        outfits = fashion_dataset.generate_outfit(item, param["topk"], param["max_outfits"])
+    unextracted_item = []
-        scores, features = service.get_result(outfits)
+    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
 
-        # print(scores)
+    # TODO 读取本次任务需要的图片特征
-        # print(len(scores))
+    for item in all_items:
-        best_outfits, best_scores = service.visualize(outfits, scores, param["topk"], best=True,
+        if item["item_name"] not in prepared_feature.keys():
-                                                      # output_path=os.path.join(r"E:\workspace\outfit_matcher\2024 SS Outfit", f"{item['item_name']}_best_{param['topk']}.png")
+            prepared_feature[item["item_name"]] = np.load(f'feature/{item["item_name"]}.npy')
-                                                      )
-        bad_outfits, bad_scores = service.visualize(outfits, scores, param["topk"], best=False,
-                                                    # output_path=os.path.join(r"E:\workspace\outfit_matcher\2024 SS Outfit", f"{item['item_name']}_worst_{param['topk']}.png")
-                                                    )
-        best_list.append({"best_outfits": best_outfits, "best_scores": best_scores})
-        bad_list.append({"bad_outfits": bad_outfits, "bad_scores": bad_scores})
 
-    pprint(best_list)
+    # 开始服装搭配任务
-    pprint(bad_list)
+    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"))