Spaces:
Sleeping
Sleeping
updated
Browse files- app.py +75 -43
- models/clip/_clip/__init__.py +0 -31
- models/clip/_clip/prepare.py +1 -8
app.py
CHANGED
@@ -40,8 +40,6 @@ truncation = 4
|
|
40 |
reduction = 8
|
41 |
granularity = "fine"
|
42 |
anchor_points = "average"
|
43 |
-
|
44 |
-
model_name = "clip_vit_l_14"
|
45 |
input_size = 224
|
46 |
|
47 |
# Comment the lines below to test non-CLIP models.
|
@@ -50,8 +48,19 @@ num_vpt = 32
|
|
50 |
vpt_drop = 0.
|
51 |
deep_vpt = True
|
52 |
|
53 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
|
|
|
55 |
|
56 |
if truncation is None: # regression, no truncation.
|
57 |
bins, anchor_points = None, None
|
@@ -62,32 +71,48 @@ else:
|
|
62 |
anchor_points = config["anchor_points"][granularity]["average"] if anchor_points == "average" else config["anchor_points"][granularity]["middle"]
|
63 |
bins = [(float(b[0]), float(b[1])) for b in bins]
|
64 |
anchor_points = [float(p) for p in anchor_points]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
65 |
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
prompt_type=prompt_type,
|
75 |
-
num_vpt=num_vpt,
|
76 |
-
vpt_drop=vpt_drop,
|
77 |
-
deep_vpt=deep_vpt
|
78 |
-
)
|
79 |
-
|
80 |
-
repo_id = "Yiming-M/CLIP-EBC"
|
81 |
-
filename = "nwpu_weights/CLIP_EBC_ViT_L_14/model.safetensors"
|
82 |
-
weights_path = hf_hub_download(repo_id, filename)
|
83 |
-
# weights_path = os.path.join("CLIP_EBC_ViT_L_14", "model.safetensors")
|
84 |
-
state_dict = load_file(weights_path)
|
85 |
-
new_state_dict = {}
|
86 |
-
for k, v in state_dict.items():
|
87 |
-
new_state_dict[k.replace("model.", "")] = v
|
88 |
-
model.load_state_dict(new_state_dict)
|
89 |
-
model.to(device)
|
90 |
-
model.eval()
|
91 |
|
92 |
|
93 |
# -----------------------------
|
@@ -114,17 +139,22 @@ def transform(image: Image.Image):
|
|
114 |
# -----------------------------
|
115 |
# Inference function
|
116 |
# -----------------------------
|
117 |
-
def predict(image: Image.Image):
|
118 |
"""
|
119 |
Given an input image, preprocess it, run the model to obtain a density map,
|
120 |
compute the total crowd count, and prepare the density map for display.
|
121 |
"""
|
|
|
|
|
|
|
|
|
|
|
122 |
# Preprocess the image
|
123 |
input_width, input_height = image.size
|
124 |
input_tensor = transform(image).to(device) # shape: (1, 3, H, W)
|
125 |
|
126 |
with torch.no_grad():
|
127 |
-
density_map =
|
128 |
total_count = density_map.sum().item()
|
129 |
resized_density_map = resize_density_map(density_map, (input_height, input_width)).cpu().squeeze().numpy()
|
130 |
|
@@ -149,32 +179,34 @@ def predict(image: Image.Image):
|
|
149 |
# Build Gradio Interface using Blocks for a two-column layout
|
150 |
# -----------------------------
|
151 |
with gr.Blocks() as demo:
|
152 |
-
gr.Markdown("# Crowd Counting
|
153 |
gr.Markdown("Upload an image or select an example below to see the predicted crowd density map and total count.")
|
154 |
-
|
155 |
with gr.Row():
|
156 |
with gr.Column():
|
157 |
-
|
158 |
-
|
159 |
-
|
160 |
-
|
161 |
)
|
|
|
162 |
submit_btn = gr.Button("Predict")
|
163 |
with gr.Column():
|
164 |
output_img = gr.Image(label="Predicted Density Map", type="pil")
|
165 |
output_text = gr.Textbox(label="Total Count")
|
166 |
-
|
167 |
-
submit_btn.click(fn=predict, inputs=input_img, outputs=[input_img, output_img, output_text])
|
168 |
-
|
169 |
-
# Optional: add example images. Ensure these files are in your repo.
|
170 |
gr.Examples(
|
171 |
examples=[
|
172 |
["example1.jpg"],
|
173 |
-
["example2.jpg"]
|
|
|
|
|
|
|
174 |
],
|
175 |
inputs=input_img,
|
176 |
label="Try an example"
|
177 |
)
|
178 |
|
179 |
-
|
180 |
-
demo.launch()
|
|
|
40 |
reduction = 8
|
41 |
granularity = "fine"
|
42 |
anchor_points = "average"
|
|
|
|
|
43 |
input_size = 224
|
44 |
|
45 |
# Comment the lines below to test non-CLIP models.
|
|
|
48 |
vpt_drop = 0.
|
49 |
deep_vpt = True
|
50 |
|
51 |
+
repo_id = "Yiming-M/CLIP-EBC"
|
52 |
+
model_configs = {
|
53 |
+
"CLIP_EBC_ViT_L_14": {
|
54 |
+
"model_name": "clip_vit_l_14",
|
55 |
+
"filename": "nwpu_weights/CLIP_EBC_ViT_L_14/model.safetensors",
|
56 |
+
},
|
57 |
+
"CLIP_EBC_ViT_B_16": {
|
58 |
+
"model_name": "clip_vit_b_16",
|
59 |
+
"filename": "nwpu_weights/CLIP_EBC_ViT_B_16/model.safetensors",
|
60 |
+
},
|
61 |
+
}
|
62 |
|
63 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
64 |
|
65 |
if truncation is None: # regression, no truncation.
|
66 |
bins, anchor_points = None, None
|
|
|
71 |
anchor_points = config["anchor_points"][granularity]["average"] if anchor_points == "average" else config["anchor_points"][granularity]["middle"]
|
72 |
bins = [(float(b[0]), float(b[1])) for b in bins]
|
73 |
anchor_points = [float(p) for p in anchor_points]
|
74 |
+
# Use a global reference to store the model instance
|
75 |
+
loaded_model = None
|
76 |
+
|
77 |
+
def load_model(model_choice: str):
|
78 |
+
global loaded_model
|
79 |
+
|
80 |
+
config = model_configs[model_choice]
|
81 |
+
model_name = config["model_name"]
|
82 |
+
filename = config["filename"]
|
83 |
+
|
84 |
+
# Prepare bins and anchor_points if using classification
|
85 |
+
if truncation is None:
|
86 |
+
bins_, anchor_points_ = None, None
|
87 |
+
else:
|
88 |
+
with open(os.path.join("configs", f"reduction_{reduction}.json"), "r") as f:
|
89 |
+
config_json = json.load(f)[str(truncation)]["nwpu"]
|
90 |
+
bins_ = config_json["bins"][granularity]
|
91 |
+
anchor_points_ = config_json["anchor_points"][granularity]["average"] if anchor_points == "average" else config_json["anchor_points"][granularity]["middle"]
|
92 |
+
bins_ = [(float(b[0]), float(b[1])) for b in bins_]
|
93 |
+
anchor_points_ = [float(p) for p in anchor_points_]
|
94 |
+
|
95 |
+
# Build model
|
96 |
+
model = get_model(
|
97 |
+
backbone=model_name,
|
98 |
+
input_size=input_size,
|
99 |
+
reduction=reduction,
|
100 |
+
bins=bins_,
|
101 |
+
anchor_points=anchor_points_,
|
102 |
+
prompt_type=prompt_type,
|
103 |
+
num_vpt=num_vpt,
|
104 |
+
vpt_drop=vpt_drop,
|
105 |
+
deep_vpt=deep_vpt,
|
106 |
+
)
|
107 |
|
108 |
+
weights_path = hf_hub_download(repo_id, filename)
|
109 |
+
state_dict = load_file(weights_path)
|
110 |
+
new_state_dict = {k.replace("model.", ""): v for k, v in state_dict.items()}
|
111 |
+
model.load_state_dict(new_state_dict)
|
112 |
+
model.to(device)
|
113 |
+
model.eval()
|
114 |
+
|
115 |
+
loaded_model = model
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
116 |
|
117 |
|
118 |
# -----------------------------
|
|
|
139 |
# -----------------------------
|
140 |
# Inference function
|
141 |
# -----------------------------
|
142 |
+
def predict(image: Image.Image, model_choice: str = "CLIP_EBC_ViT_B_16"):
|
143 |
"""
|
144 |
Given an input image, preprocess it, run the model to obtain a density map,
|
145 |
compute the total crowd count, and prepare the density map for display.
|
146 |
"""
|
147 |
+
global loaded_model
|
148 |
+
|
149 |
+
if loaded_model is None or model_configs[model_choice]["model_name"] not in loaded_model.__class__.__name__:
|
150 |
+
load_model(model_choice)
|
151 |
+
|
152 |
# Preprocess the image
|
153 |
input_width, input_height = image.size
|
154 |
input_tensor = transform(image).to(device) # shape: (1, 3, H, W)
|
155 |
|
156 |
with torch.no_grad():
|
157 |
+
density_map = loaded_model(input_tensor) # expected shape: (1, 1, H, W)
|
158 |
total_count = density_map.sum().item()
|
159 |
resized_density_map = resize_density_map(density_map, (input_height, input_width)).cpu().squeeze().numpy()
|
160 |
|
|
|
179 |
# Build Gradio Interface using Blocks for a two-column layout
|
180 |
# -----------------------------
|
181 |
with gr.Blocks() as demo:
|
182 |
+
gr.Markdown("# Crowd Counting by CLIP-EBC (Pre-trained on NWPU-Crowd)")
|
183 |
gr.Markdown("Upload an image or select an example below to see the predicted crowd density map and total count.")
|
184 |
+
|
185 |
with gr.Row():
|
186 |
with gr.Column():
|
187 |
+
model_choice = gr.Dropdown(
|
188 |
+
choices=list(model_configs.keys()),
|
189 |
+
value="CLIP_EBC_ViT_B_16",
|
190 |
+
label="Select Model"
|
191 |
)
|
192 |
+
input_img = gr.Image(label="Input Image", sources=["upload", "clipboard"], type="pil")
|
193 |
submit_btn = gr.Button("Predict")
|
194 |
with gr.Column():
|
195 |
output_img = gr.Image(label="Predicted Density Map", type="pil")
|
196 |
output_text = gr.Textbox(label="Total Count")
|
197 |
+
|
198 |
+
submit_btn.click(fn=predict, inputs=[input_img, model_choice], outputs=[input_img, output_img, output_text])
|
199 |
+
|
|
|
200 |
gr.Examples(
|
201 |
examples=[
|
202 |
["example1.jpg"],
|
203 |
+
["example2.jpg"],
|
204 |
+
["example3.jpg"],
|
205 |
+
["example4.jpg"],
|
206 |
+
["example5.jpg"],
|
207 |
],
|
208 |
inputs=input_img,
|
209 |
label="Try an example"
|
210 |
)
|
211 |
|
212 |
+
demo.launch()
|
|
models/clip/_clip/__init__.py
CHANGED
@@ -13,15 +13,8 @@ from .model import CLIP
|
|
13 |
curr_dir = os.path.dirname(os.path.abspath(__file__))
|
14 |
|
15 |
clip_model_names = [
|
16 |
-
"clip_resnet50",
|
17 |
-
"clip_resnet101",
|
18 |
-
"clip_resnet50x4",
|
19 |
-
"clip_resnet50x16",
|
20 |
-
"clip_resnet50x64",
|
21 |
-
"clip_vit_b_32",
|
22 |
"clip_vit_b_16",
|
23 |
"clip_vit_l_14",
|
24 |
-
"clip_vit_l_14_336px",
|
25 |
]
|
26 |
|
27 |
clip_image_encoder_names = [f"clip_image_encoder_{name[5:]}" for name in clip_model_names]
|
@@ -240,34 +233,10 @@ __all__ = [
|
|
240 |
# utils
|
241 |
"tokenize",
|
242 |
"transform",
|
243 |
-
# clip models
|
244 |
-
"resnet50_clip",
|
245 |
-
"resnet101_clip",
|
246 |
-
"resnet50x4_clip",
|
247 |
-
"resnet50x16_clip",
|
248 |
-
"resnet50x64_clip",
|
249 |
-
"vit_b_32_clip",
|
250 |
-
"vit_b_16_clip",
|
251 |
-
"vit_l_14_clip",
|
252 |
-
"vit_l_14_336px_clip",
|
253 |
# clip image encoders
|
254 |
-
"resnet50_img",
|
255 |
-
"resnet101_img",
|
256 |
-
"resnet50x4_img",
|
257 |
-
"resnet50x16_img",
|
258 |
-
"resnet50x64_img",
|
259 |
-
"vit_b_32_img",
|
260 |
"vit_b_16_img",
|
261 |
"vit_l_14_img",
|
262 |
-
"vit_l_14_336px_img",
|
263 |
# clip text encoders
|
264 |
-
"resnet50_txt",
|
265 |
-
"resnet101_txt",
|
266 |
-
"resnet50x4_txt",
|
267 |
-
"resnet50x16_txt",
|
268 |
-
"resnet50x64_txt",
|
269 |
-
"vit_b_32_txt",
|
270 |
"vit_b_16_txt",
|
271 |
"vit_l_14_txt",
|
272 |
-
"vit_l_14_336px_txt",
|
273 |
]
|
|
|
13 |
curr_dir = os.path.dirname(os.path.abspath(__file__))
|
14 |
|
15 |
clip_model_names = [
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
"clip_vit_b_16",
|
17 |
"clip_vit_l_14",
|
|
|
18 |
]
|
19 |
|
20 |
clip_image_encoder_names = [f"clip_image_encoder_{name[5:]}" for name in clip_model_names]
|
|
|
233 |
# utils
|
234 |
"tokenize",
|
235 |
"transform",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
236 |
# clip image encoders
|
|
|
|
|
|
|
|
|
|
|
|
|
237 |
"vit_b_16_img",
|
238 |
"vit_l_14_img",
|
|
|
239 |
# clip text encoders
|
|
|
|
|
|
|
|
|
|
|
|
|
240 |
"vit_b_16_txt",
|
241 |
"vit_l_14_txt",
|
|
|
242 |
]
|
models/clip/_clip/prepare.py
CHANGED
@@ -9,15 +9,8 @@ from .utils import load
|
|
9 |
|
10 |
|
11 |
model_name_map = {
|
12 |
-
"RN50": "resnet50",
|
13 |
-
"RN101": "resnet101",
|
14 |
-
"RN50x4": "resnet50x4",
|
15 |
-
"RN50x16": "resnet50x16",
|
16 |
-
"RN50x64": "resnet50x64",
|
17 |
-
"ViT-B/32": "vit_b_32",
|
18 |
"ViT-B/16": "vit_b_16",
|
19 |
"ViT-L/14": "vit_l_14",
|
20 |
-
"ViT-L/14@336px": "vit_l_14_336px",
|
21 |
}
|
22 |
|
23 |
|
@@ -49,7 +42,7 @@ def prepare() -> None:
|
|
49 |
os.makedirs(config_dir, exist_ok=True)
|
50 |
device = torch.device("cpu")
|
51 |
|
52 |
-
for model_name in tqdm(["
|
53 |
model = load(model_name, device=device).to(device)
|
54 |
image_encoder = model.visual.to(device)
|
55 |
text_encoder = CLIPTextEncoderTemp(model).to(device)
|
|
|
9 |
|
10 |
|
11 |
model_name_map = {
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
"ViT-B/16": "vit_b_16",
|
13 |
"ViT-L/14": "vit_l_14",
|
|
|
14 |
}
|
15 |
|
16 |
|
|
|
42 |
os.makedirs(config_dir, exist_ok=True)
|
43 |
device = torch.device("cpu")
|
44 |
|
45 |
+
for model_name in tqdm(["ViT-B/16", "ViT-L/14"]):
|
46 |
model = load(model_name, device=device).to(device)
|
47 |
image_encoder = model.visual.to(device)
|
48 |
text_encoder = CLIPTextEncoderTemp(model).to(device)
|