File size: 8,405 Bytes
5e0e1ed 8b7c9d4 5e0e1ed 8b7c9d4 5e0e1ed 8b7c9d4 5e0e1ed 8b7c9d4 5e0e1ed 8b7c9d4 5e0e1ed | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 | import gradio as gr
import torch
import numpy as np
import cv2
import tempfile
from PIL import Image
import torchvision.transforms as transforms
import os
from huggingface_hub import hf_hub_download
import huggingface_hub
from models.pvsnet_model import PVSNet
import helperFunctions as helper
import parameters as params
DEVICE = params.DEVICE
def getPositionVector(x, y, z, pose_dims=3):
if pose_dims == 3:
vector = torch.zeros((1, 3), dtype=torch.float)
vector[0][0] = (float(format(x, '.7f')) - (-0.1)) / (0.1 - (-0.1))
vector[0][1] = (float(format(y, '.7f')) - (-0.1)) / (0.1 - (-0.1))
vector[0][2] = (float(format(z, '.7f')) - (-0.1)) / (0.1 - (-0.1))
return vector
else:
t_min, t_max = -0.1, 0.1
r_min, r_max = -3, 3
vector = torch.zeros((1, 6), dtype=torch.float)
vector[0, 0] = (x - t_min) / (t_max - t_min)
vector[0, 1] = (y - t_min) / (t_max - t_min)
vector[0, 2] = (z - t_min) / (t_max - t_min)
vector[0, 3] = (0 - r_min) / (r_max - r_min)
vector[0, 4] = (0 - r_min) / (r_max - r_min)
vector[0, 5] = (0 - r_min) / (r_max - r_min)
return vector
def generateCircularTrajectory(radius, num_frames):
angles = np.linspace(0, 2 * np.pi, num_frames, endpoint=False)
return [[radius * np.cos(angle), radius * np.sin(angle), 0] for angle in angles]
def generateSwingTrajectory(radius, num_frames):
angles = np.linspace(0, 2 * np.pi, num_frames, endpoint=False)
return [[radius * np.cos(angle), 0, radius * np.sin(angle)] for angle in angles]
def create_video_from_memory(frames, fps=30):
if not frames:
return None
height, width, _ = frames[0].shape
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
temp_video = tempfile.NamedTemporaryFile(delete=False, suffix=".mp4")
out = cv2.VideoWriter(temp_video.name, fourcc, fps, (width, height))
for frame in frames:
out.write(frame)
out.release()
return temp_video.name
def process_image(img, video_type, radius, num_frames, num_loops, dataset, resolution, architecture):
if img is None:
return None
width, height = map(int, resolution.split('x'))
min_dim = min(img.width, img.height)
left = (img.width - min_dim) / 2
top = (img.height - min_dim) / 2
right = (img.width + min_dim) / 2
bottom = (img.height + min_dim) / 2
img = img.crop((left, top, right, bottom))
is_lite = (architecture == "Lite")
pose_dims = 3 if dataset == "Blender" else 6
dataset_prefix = dataset.lower()
arch_infix = "pvsnet_lite" if is_lite else "pvsnet"
checkpoint_name = f"checkpoint_{dataset_prefix}_{arch_infix}_{resolution}.pth"
try:
checkpoint_path = hf_hub_download(
repo_id="3ZadeSSG/PVSNet",
filename=checkpoint_name
)
except huggingface_hub.utils.EntryNotFoundError:
raise gr.Error(f"Checkpoint {checkpoint_name} not found in Hugging Face Hub! Please select a valid combination.")
except Exception as e:
raise gr.Error(f"Error downloading checkpoint {checkpoint_name}: {e}")
model = PVSNet(total_image_input=params.params_number_input, pose_dims=pose_dims, height=height, width=width, is_lite=is_lite)
try:
model = helper.load_Checkpoint(checkpoint_path, model, load_cpu=True)
except Exception as e:
print(f"Error loading checkpoint {checkpoint_path}: {e}")
raise gr.Error(f"Error loading checkpoint {checkpoint_path}: {e}")
model.to(DEVICE)
model.eval()
transform = transforms.Compose([
transforms.Resize((height, width)),
transforms.ToTensor()
])
img_input = img.convert('RGB')
img_input = transform(img_input).unsqueeze(0).to(DEVICE)
if video_type == "Circle":
raw_traj = generateCircularTrajectory(radius, num_frames)
trajectory = [(p[0], p[1], 0) for p in raw_traj]
elif video_type == "Swing":
raw_traj = generateSwingTrajectory(radius, num_frames)
trajectory = raw_traj
else:
raw_traj = generateCircularTrajectory(radius, num_frames)
trajectory = [(p[0], p[1], 0) for p in raw_traj]
view_frames = []
for x, y, z in trajectory:
pos = getPositionVector(x, y, z, pose_dims=pose_dims).unsqueeze(0).to(DEVICE)
with torch.no_grad():
predicted_img = model(img_input, pos)
p_img = predicted_img[0].detach().cpu().permute(1, 2, 0).numpy()
p_img = np.clip(p_img, 0, 1)
p_img = (p_img * 255).astype(np.uint8)
p_img_bgr = cv2.cvtColor(p_img, cv2.COLOR_RGB2BGR)
view_frames.append(p_img_bgr)
view_frames = view_frames * int(num_loops)
fps = 60
view_video_path = create_video_from_memory(view_frames, fps=fps)
return view_video_path
with gr.Blocks(title="PVSNet", theme="default") as demo:
gr.Markdown(
"""
## PVSNet: Real-Time Position-Aware View Synthesis from Single-View Input
* Upload an image and get a mini video showing capability of novel view synthesis.
**Note:** Huggingface demo is running on CPU so inference speeds will be slow. Inference might take around 2-5 mins depending on resolution and model. We recomment runnning it on 256x256 and with Lite model.
### Head to our [Project Page](https://realistic3d-miun.github.io/PVSNet/) for more details about the models
""")
with gr.Row():
with gr.Column():
img_input = gr.Image(type="pil", label="Input Image", height=256)
with gr.Group():
dataset_type = gr.Dropdown(["Blender", "COCO"], label="Dataset Model", value="COCO")
resolution_type = gr.Dropdown(["256x256", "512x512"], label="Resolution", value="256x256")
architecture_type = gr.Dropdown(["Regular", "Lite"], label="Architecture", value="Lite")
video_type = gr.Dropdown(["Circle", "Swing"], label="Trajectory Type", value="Swing")
with gr.Accordion("Advanced Settings", open=False):
radius = gr.Slider(0.01, 0.1, value=0.06, label="Motion Radius")
num_frames = gr.Slider(10, 120, value=60, step=1, label="Frames per Loop")
num_loops = gr.Slider(1, 6, value=3, step=1, label="Number of Loops")
submit_btn = gr.Button("Generate", variant="primary")
with gr.Column():
video_output = gr.Video(label="Generated View Video", height=256)
submit_btn.click(
fn=process_image,
inputs=[img_input, video_type, radius, num_frames, num_loops, dataset_type, resolution_type, architecture_type],
outputs=[video_output]
)
gr.Markdown("### Example Images: Click to Load")
import glob
blender_imgs = sorted(glob.glob("./sample_images/blender/*"))
coco_imgs = sorted(glob.glob("./sample_images/coco/*"))
rw_imgs = sorted(glob.glob("./sample_images/real_world/*"))
def create_grid(imgs, title, cols=4):
gr.Markdown(title)
image_components = []
for i in range(0, len(imgs), cols):
with gr.Row():
for img_path in imgs[i:i+cols]:
comp = gr.Image(img_path, label=os.path.basename(img_path), height=150, interactive=False, show_label=True)
image_components.append((comp, img_path))
return image_components
with gr.Column():
b_comps = create_grid(blender_imgs, "#### Blender Models (Loads Blender Lite 256x256 by Default)")
c_comps = create_grid(coco_imgs, "#### COCO Models (Loads COCO 256x256 Lite by Default)")
r_comps = create_grid(rw_imgs, "#### Real World Models (Loads COCO 256x256 Lite by Default)")
for comp, path in b_comps:
comp.select(fn=lambda p=path: (Image.open(p), "Blender", "256x256", "Lite"), outputs=[img_input, dataset_type, resolution_type, architecture_type])
for comp, path in c_comps:
comp.select(fn=lambda p=path: (Image.open(p), "COCO", "256x256", "Lite"), outputs=[img_input, dataset_type, resolution_type, architecture_type])
for comp, path in r_comps:
comp.select(fn=lambda p=path: (Image.open(p), "COCO", "256x256", "Lite"), outputs=[img_input, dataset_type, resolution_type, architecture_type])
if __name__ == "__main__":
demo.launch()
|