feat: faster rcnn generate bouding box's json

src/test_backbone.py

@@ -0,0 +1,44 @@
+import torch
+import torchvision
+import torchvision.models as models
+import time
+from torchinfo import summary
+import torchvision.transforms as transforms
+from PIL import Image
+import matplotlib.pyplot as plt
+
+device = torch.device('cuda')
+
+# 創建 resnet50 模型
+model = models.resnet50(weights=torchvision.models.ResNet50_Weights).to(device)
+model = torch.nn.Sequential(*list(model.children())[:-3])                           # 去除 FC & global pooling  && 一層
+
+# 定義 transform
+transform = transforms.Compose([
+    transforms.Resize((224, 224)),
+    transforms.ToTensor(),
+    transforms.Normalize(mean=[0.485, 0.456, 0.406], 
+                         std=[0.229, 0.224, 0.225])
+])
+
+
+for name, param in model.named_parameters():
+    param.requires_grad = False
+    # print(name, param.requires_grad)
+# 將模型設置為 evaluation 模式
+model.eval()
+
+# summary(model, input_size=(32, 3, 224, 224))
+# time.sleep(2)
+
+with torch.no_grad():
+    image = Image.open('../data/textcaps/train_images/{}'.format('0a9d5e54fcf25702.jpg'))
+    image = transform(image).to(device).unsqueeze(0)
+    
+    output = model(image)
+    output = output[0].to('cpu')
+    for index, feature in enumerate(output):
+        plt.imshow(feature)
+        plt.savefig("result/tmp/{}.jpg".format(index))
+    print(output.shape)
+# print(model)

src/test_object_detection.py

@@ -0,0 +1,85 @@
+import torchvision
+import torch
+import torchvision.transforms as transforms
+from torchvision.models.detection import fasterrcnn_resnet50_fpn, FasterRCNN_ResNet50_FPN_Weights
+import os
+import json
+from PIL import Image
+import matplotlib.pyplot as plt
+import matplotlib.patches as patches
+
+device = torch.device('cuda')
+
+classes = {1: 'person', 2: 'bicycle', 3: 'car', 4: 'motorcycle', 5: 'airplane', 6: 'bus', 7: 'train', 8: 'truck', 9: 'boat', 10: 'traffic light', 11: 'fire hydrant', 13: 'stop sign', 14: 'parking meter', 15: 'bench', 16: 'bird', 17: 'cat', 18: 'dog', 19: 'horse', 20: 'sheep', 21: 'cow', 22: 'elephant', 23: 'bear', 24: 'zebra', 25: 'giraffe', 27: 'backpack', 28: 'umbrella', 31: 'handbag', 32: 'tie', 33: 'suitcase', 34: 'frisbee', 35: 'skis', 36: 'snowboard', 37: 'sports ball', 38: 'kite', 39: 'baseball bat', 40: 'baseball glove', 41: 'skateboard', 42: 'surfboard', 43: 'tennis racket', 44: 'bottle', 46: 'wine glass', 47: 'cup', 48: 'fork', 49: 'knife', 50: 'spoon', 51: 'bowl', 52: 'banana', 53: 'apple', 54: 'sandwich', 55: 'orange', 56: 'broccoli', 57: 'carrot', 58: 'hot dog', 59: 'pizza', 60: 'donut', 61: 'cake', 62: 'chair', 63: 'couch', 64: 'potted plant', 65: 'bed', 67: 'dining table', 70: 'toilet', 72: 'tv', 73: 'laptop', 74: 'mouse', 75: 'remote', 76: 'keyboard', 77: 'cell phone', 78: 'microwave', 79: 'oven', 80: 'toaster', 81: 'sink', 82: 'refrigerator', 84: 'book', 85: 'clock', 86: 'vase', 87: 'scissors', 88: 'teddy bear', 89: 'hair drier', 90: 'toothbrush'}
+
+# classes is from here
+#
+#
+# with open('../data/coco/annotations/instances_train2017.json') as fp:
+#     data = json.load(fp)
+# for i in data['categories']:
+#     classes[i['id']] = i['name']
+# print(classes)
+
+model = fasterrcnn_resnet50_fpn(weights=FasterRCNN_ResNet50_FPN_Weights.COCO_V1).to(device)
+model.eval()
+
+
+
+
+# 定義轉換
+transform = transforms.Compose([
+    transforms.Resize((256, 256)),
+    transforms.ToTensor()
+])
+
+images = []
+paths = []
+count = 0
+l = len(os.listdir('../data/textcaps/train_images'))
+
+for filename in os.listdir("../data/textcaps/train_images"):
+    if len(images) < 32:
+        image = Image.open('../data/textcaps/train_images/{}'.format(filename))
+        image = transform(image).to(device)
+        c, h, w = image.shape
+        if c == 3:
+            images.append(image)
+            paths.append(filename)
+    else:
+        with torch.no_grad():
+            outputs = model(images)
+        
+        for fileindex, output in enumerate(outputs):
+
+            # fig, ax = plt.subplots(1)
+            # image = images[fileindex].to('cpu')
+            # ax.imshow(image.permute(1, 2, 0))
+
+            datas = {}
+            datas['filename'] = paths[fileindex]
+            datas['boxes'] = []
+
+            box_counter = 0
+            for index in range(len(output['boxes'])):
+                box = output['boxes'][index].tolist()
+                score = output['scores'][index].item()
+                label = output['labels'][index].item()
+                if score > 0.3:  # 只畫出信心度大於 0.5 的 bounding box
+                    # rect = patches.Rectangle((box[0], box[1]), box[2]-box[0], box[3]-box[1], linewidth=2, edgecolor='r', facecolor='none')
+                    # ax.add_patch(rect)
+                    # ax.text(box[0], box[1]-5, classes[label] + f' {score:.2f}', color='r', fontsize=12)
+                    box_counter += 1
+                    datas['boxes'].append({'box': box, 'label': classes[label], 'score': score})
+            # print(datas)
+            # if box_counter < 10:
+            #     plt.savefig("result/{}".format(paths[fileindex]))
+            with open('result/object_detection/{}.json'.format(paths[fileindex]), 'w') as fp:
+                json.dump(datas, fp)
+            
+        images = []
+        paths = []
+        print("batch finished.")
+    count += 1
+    print("{}/{}".format(count, l))
+