Spaces:
Runtime error
Runtime error
RamAnanth1
commited on
Commit
•
6585503
1
Parent(s):
e067e6a
Update app.py
Browse filesAdd function to include human pose in control tasks
app.py
CHANGED
@@ -45,6 +45,8 @@ def process(input_image, prompt, input_control, a_prompt, n_prompt, num_samples,
|
|
45 |
# TODO: Add other control tasks
|
46 |
if input_control == "Scribble":
|
47 |
return process_scribble(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta)
|
|
|
|
|
48 |
return process_canny(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta, low_threshold, high_threshold)
|
49 |
|
50 |
def process_canny(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta, low_threshold, high_threshold):
|
@@ -104,6 +106,42 @@ def process_scribble(input_image, prompt, a_prompt, n_prompt, num_samples, image
|
|
104 |
|
105 |
results = [x_samples[i] for i in range(num_samples)]
|
106 |
return [255 - detected_map] + results
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
107 |
|
108 |
def create_canvas(w, h):
|
109 |
new_control_options = ["Interactive Scribble"]
|
@@ -113,7 +151,8 @@ def create_canvas(w, h):
|
|
113 |
block = gr.Blocks().queue()
|
114 |
control_task_list = [
|
115 |
"Canny Edge Map",
|
116 |
-
"Scribble"
|
|
|
117 |
]
|
118 |
with block:
|
119 |
gr.Markdown("## Adding Conditional Control to Text-to-Image Diffusion Models")
|
|
|
45 |
# TODO: Add other control tasks
|
46 |
if input_control == "Scribble":
|
47 |
return process_scribble(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta)
|
48 |
+
elif input_control == "Pose":
|
49 |
+
return process_pose(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, image_resolution, ddim_steps, scale, seed, eta)
|
50 |
return process_canny(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta, low_threshold, high_threshold)
|
51 |
|
52 |
def process_canny(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, ddim_steps, scale, seed, eta, low_threshold, high_threshold):
|
|
|
106 |
|
107 |
results = [x_samples[i] for i in range(num_samples)]
|
108 |
return [255 - detected_map] + results
|
109 |
+
|
110 |
+
def process_pose(input_image, prompt, a_prompt, n_prompt, num_samples, image_resolution, detect_resolution, ddim_steps, scale, seed, eta):
|
111 |
+
with torch.no_grad():
|
112 |
+
input_image = HWC3(input_image)
|
113 |
+
detected_map, _ = apply_openpose(resize_image(input_image, detect_resolution))
|
114 |
+
detected_map = HWC3(detected_map)
|
115 |
+
img = resize_image(input_image, image_resolution)
|
116 |
+
H, W, C = img.shape
|
117 |
+
|
118 |
+
detected_map = cv2.resize(detected_map, (W, H), interpolation=cv2.INTER_NEAREST)
|
119 |
+
|
120 |
+
control = torch.from_numpy(detected_map.copy()).float().cuda() / 255.0
|
121 |
+
control = torch.stack([control for _ in range(num_samples)], dim=0)
|
122 |
+
control = einops.rearrange(control, 'b h w c -> b c h w').clone()
|
123 |
+
|
124 |
+
if seed == -1:
|
125 |
+
seed = random.randint(0, 65535)
|
126 |
+
seed_everything(seed)
|
127 |
+
|
128 |
+
|
129 |
+
cond = {"c_concat": [control], "c_crossattn": [pose_model.get_learned_conditioning([prompt + ', ' + a_prompt] * num_samples)]}
|
130 |
+
un_cond = {"c_concat": [control], "c_crossattn": [pose_model.get_learned_conditioning([n_prompt] * num_samples)]}
|
131 |
+
shape = (4, H // 8, W // 8)
|
132 |
+
|
133 |
+
|
134 |
+
samples, intermediates = ddim_sampler_pose.sample(ddim_steps, num_samples,
|
135 |
+
shape, cond, verbose=False, eta=eta,
|
136 |
+
unconditional_guidance_scale=scale,
|
137 |
+
unconditional_conditioning=un_cond)
|
138 |
+
|
139 |
+
|
140 |
+
x_samples = pose_model.decode_first_stage(samples)
|
141 |
+
x_samples = (einops.rearrange(x_samples, 'b c h w -> b h w c') * 127.5 + 127.5).cpu().numpy().clip(0, 255).astype(np.uint8)
|
142 |
+
|
143 |
+
results = [x_samples[i] for i in range(num_samples)]
|
144 |
+
return [detected_map] + results
|
145 |
|
146 |
def create_canvas(w, h):
|
147 |
new_control_options = ["Interactive Scribble"]
|
|
|
151 |
block = gr.Blocks().queue()
|
152 |
control_task_list = [
|
153 |
"Canny Edge Map",
|
154 |
+
"Scribble",
|
155 |
+
"Pose"
|
156 |
]
|
157 |
with block:
|
158 |
gr.Markdown("## Adding Conditional Control to Text-to-Image Diffusion Models")
|