try without partial
Browse files
app.py
CHANGED
|
@@ -10,16 +10,13 @@ import sys
|
|
| 10 |
import os.path as path
|
| 11 |
import torch
|
| 12 |
import tempfile
|
|
|
|
| 13 |
|
| 14 |
HERE_PATH = path.normpath(path.dirname(__file__)) # noqa
|
| 15 |
MASt3R_REPO_PATH = path.normpath(path.join(HERE_PATH, './mast3r')) # noqa
|
| 16 |
sys.path.insert(0, MASt3R_REPO_PATH) # noqa
|
| 17 |
|
| 18 |
-
|
| 19 |
-
mast3r.demo.get_reconstructed_scene = spaces.GPU(mast3r.demo.get_reconstructed_scene)
|
| 20 |
-
mast3r.demo.get_3D_model_from_scene = spaces.GPU(mast3r.demo.get_3D_model_from_scene)
|
| 21 |
-
|
| 22 |
-
from mast3r.demo import main_demo
|
| 23 |
from mast3r.model import AsymmetricMASt3R
|
| 24 |
from mast3r.utils.misc import hash_md5
|
| 25 |
|
|
@@ -35,9 +32,138 @@ device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
|
| 35 |
model = AsymmetricMASt3R.from_pretrained(weights_path).to(device)
|
| 36 |
chkpt_tag = hash_md5(weights_path)
|
| 37 |
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
import os.path as path
|
| 11 |
import torch
|
| 12 |
import tempfile
|
| 13 |
+
import gradio
|
| 14 |
|
| 15 |
HERE_PATH = path.normpath(path.dirname(__file__)) # noqa
|
| 16 |
MASt3R_REPO_PATH = path.normpath(path.join(HERE_PATH, './mast3r')) # noqa
|
| 17 |
sys.path.insert(0, MASt3R_REPO_PATH) # noqa
|
| 18 |
|
| 19 |
+
from mast3r.demo import get_reconstructed_scene, get_3D_model_from_scene, set_scenegraph_options
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
from mast3r.model import AsymmetricMASt3R
|
| 21 |
from mast3r.utils.misc import hash_md5
|
| 22 |
|
|
|
|
| 32 |
model = AsymmetricMASt3R.from_pretrained(weights_path).to(device)
|
| 33 |
chkpt_tag = hash_md5(weights_path)
|
| 34 |
|
| 35 |
+
tmpdirname = "tmp/gradio"
|
| 36 |
+
image_size = 512
|
| 37 |
+
silent = True
|
| 38 |
+
gradio_delete_cache = 7200
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
@spaces.GPU()
|
| 42 |
+
def local_get_reconstructed_scene(current_scene_state,
|
| 43 |
+
filelist, optim_level, lr1, niter1, lr2, niter2, min_conf_thr, matching_conf_thr,
|
| 44 |
+
as_pointcloud, mask_sky, clean_depth, transparent_cams, cam_size, scenegraph_type, winsize,
|
| 45 |
+
win_cyclic, refid, TSDF_thresh, shared_intrinsics, **kw):
|
| 46 |
+
return get_reconstructed_scene(tmpdirname, gradio_delete_cache, model, device, silent, image_size, current_scene_state,
|
| 47 |
+
filelist, optim_level, lr1, niter1, lr2, niter2, min_conf_thr, matching_conf_thr,
|
| 48 |
+
as_pointcloud, mask_sky, clean_depth, transparent_cams, cam_size, scenegraph_type, winsize,
|
| 49 |
+
win_cyclic, refid, TSDF_thresh, shared_intrinsics, **kw)
|
| 50 |
+
|
| 51 |
+
|
| 52 |
+
@spaces.GPU()
|
| 53 |
+
def local_get_3D_model_from_scene(scene_state, min_conf_thr=2, as_pointcloud=False, mask_sky=False,
|
| 54 |
+
clean_depth=False, transparent_cams=False, cam_size=0.05, TSDF_thresh=0):
|
| 55 |
+
return get_3D_model_from_scene(silent, scene_state, min_conf_thr, as_pointcloud, mask_sky,
|
| 56 |
+
clean_depth, transparent_cams, cam_size, TSDF_thresh)
|
| 57 |
+
|
| 58 |
+
|
| 59 |
+
recon_fun = local_get_reconstructed_scene
|
| 60 |
+
model_from_scene_fun = local_get_3D_model_from_scene
|
| 61 |
+
|
| 62 |
+
|
| 63 |
+
def get_context(delete_cache):
|
| 64 |
+
css = """.gradio-container {margin: 0 !important; min-width: 100%};"""
|
| 65 |
+
title = "MASt3R Demo"
|
| 66 |
+
if delete_cache:
|
| 67 |
+
return gradio.Blocks(css=css, title=title, delete_cache=(delete_cache, delete_cache))
|
| 68 |
+
else:
|
| 69 |
+
return gradio.Blocks(css=css, title="MASt3R Demo") # for compatibility with older versions
|
| 70 |
+
|
| 71 |
+
|
| 72 |
+
with get_context(gradio_delete_cache) as demo:
|
| 73 |
+
# scene state is save so that you can change conf_thr, cam_size... without rerunning the inference
|
| 74 |
+
scene = gradio.State(None)
|
| 75 |
+
gradio.HTML('<h2 style="text-align: center;">MASt3R Demo</h2>')
|
| 76 |
+
with gradio.Column():
|
| 77 |
+
inputfiles = gradio.File(file_count="multiple")
|
| 78 |
+
with gradio.Row():
|
| 79 |
+
with gradio.Column():
|
| 80 |
+
with gradio.Row():
|
| 81 |
+
lr1 = gradio.Slider(label="Coarse LR", value=0.07, minimum=0.01, maximum=0.2, step=0.01)
|
| 82 |
+
niter1 = gradio.Number(value=500, precision=0, minimum=0, maximum=10_000,
|
| 83 |
+
label="num_iterations", info="For coarse alignment!")
|
| 84 |
+
lr2 = gradio.Slider(label="Fine LR", value=0.014, minimum=0.005, maximum=0.05, step=0.001)
|
| 85 |
+
niter2 = gradio.Number(value=200, precision=0, minimum=0, maximum=100_000,
|
| 86 |
+
label="num_iterations", info="For refinement!")
|
| 87 |
+
optim_level = gradio.Dropdown(["coarse", "refine", "refine+depth"],
|
| 88 |
+
value='refine', label="OptLevel",
|
| 89 |
+
info="Optimization level")
|
| 90 |
+
with gradio.Row():
|
| 91 |
+
matching_conf_thr = gradio.Slider(label="Matching Confidence Thr", value=5.,
|
| 92 |
+
minimum=0., maximum=30., step=0.1,
|
| 93 |
+
info="Before Fallback to Regr3D!")
|
| 94 |
+
shared_intrinsics = gradio.Checkbox(value=False, label="Shared intrinsics",
|
| 95 |
+
info="Only optimize one set of intrinsics for all views")
|
| 96 |
+
scenegraph_type = gradio.Dropdown([("complete: all possible image pairs", "complete"),
|
| 97 |
+
("swin: sliding window", "swin"),
|
| 98 |
+
("logwin: sliding window with long range", "logwin"),
|
| 99 |
+
("oneref: match one image with all", "oneref")],
|
| 100 |
+
value='complete', label="Scenegraph",
|
| 101 |
+
info="Define how to make pairs",
|
| 102 |
+
interactive=True)
|
| 103 |
+
with gradio.Column(visible=False) as win_col:
|
| 104 |
+
winsize = gradio.Slider(label="Scene Graph: Window Size", value=1,
|
| 105 |
+
minimum=1, maximum=1, step=1)
|
| 106 |
+
win_cyclic = gradio.Checkbox(value=False, label="Cyclic sequence")
|
| 107 |
+
refid = gradio.Slider(label="Scene Graph: Id", value=0,
|
| 108 |
+
minimum=0, maximum=0, step=1, visible=False)
|
| 109 |
+
run_btn = gradio.Button("Run")
|
| 110 |
+
|
| 111 |
+
with gradio.Row():
|
| 112 |
+
# adjust the confidence threshold
|
| 113 |
+
min_conf_thr = gradio.Slider(label="min_conf_thr", value=1.5, minimum=0.0, maximum=10, step=0.1)
|
| 114 |
+
# adjust the camera size in the output pointcloud
|
| 115 |
+
cam_size = gradio.Slider(label="cam_size", value=0.2, minimum=0.001, maximum=1.0, step=0.001)
|
| 116 |
+
TSDF_thresh = gradio.Slider(label="TSDF Threshold", value=0., minimum=0., maximum=1., step=0.01)
|
| 117 |
+
with gradio.Row():
|
| 118 |
+
as_pointcloud = gradio.Checkbox(value=True, label="As pointcloud")
|
| 119 |
+
# two post process implemented
|
| 120 |
+
mask_sky = gradio.Checkbox(value=False, label="Mask sky")
|
| 121 |
+
clean_depth = gradio.Checkbox(value=True, label="Clean-up depthmaps")
|
| 122 |
+
transparent_cams = gradio.Checkbox(value=False, label="Transparent cameras")
|
| 123 |
+
|
| 124 |
+
outmodel = gradio.Model3D()
|
| 125 |
+
|
| 126 |
+
# events
|
| 127 |
+
scenegraph_type.change(set_scenegraph_options,
|
| 128 |
+
inputs=[inputfiles, win_cyclic, refid, scenegraph_type],
|
| 129 |
+
outputs=[win_col, winsize, win_cyclic, refid])
|
| 130 |
+
inputfiles.change(set_scenegraph_options,
|
| 131 |
+
inputs=[inputfiles, win_cyclic, refid, scenegraph_type],
|
| 132 |
+
outputs=[win_col, winsize, win_cyclic, refid])
|
| 133 |
+
win_cyclic.change(set_scenegraph_options,
|
| 134 |
+
inputs=[inputfiles, win_cyclic, refid, scenegraph_type],
|
| 135 |
+
outputs=[win_col, winsize, win_cyclic, refid])
|
| 136 |
+
run_btn.click(fn=recon_fun,
|
| 137 |
+
inputs=[scene, inputfiles, optim_level, lr1, niter1, lr2, niter2, min_conf_thr, matching_conf_thr,
|
| 138 |
+
as_pointcloud, mask_sky, clean_depth, transparent_cams, cam_size,
|
| 139 |
+
scenegraph_type, winsize, win_cyclic, refid, TSDF_thresh, shared_intrinsics],
|
| 140 |
+
outputs=[scene, outmodel])
|
| 141 |
+
min_conf_thr.release(fn=model_from_scene_fun,
|
| 142 |
+
inputs=[scene, min_conf_thr, as_pointcloud, mask_sky,
|
| 143 |
+
clean_depth, transparent_cams, cam_size, TSDF_thresh],
|
| 144 |
+
outputs=outmodel)
|
| 145 |
+
cam_size.change(fn=model_from_scene_fun,
|
| 146 |
+
inputs=[scene, min_conf_thr, as_pointcloud, mask_sky,
|
| 147 |
+
clean_depth, transparent_cams, cam_size, TSDF_thresh],
|
| 148 |
+
outputs=outmodel)
|
| 149 |
+
TSDF_thresh.change(fn=model_from_scene_fun,
|
| 150 |
+
inputs=[scene, min_conf_thr, as_pointcloud, mask_sky,
|
| 151 |
+
clean_depth, transparent_cams, cam_size, TSDF_thresh],
|
| 152 |
+
outputs=outmodel)
|
| 153 |
+
as_pointcloud.change(fn=model_from_scene_fun,
|
| 154 |
+
inputs=[scene, min_conf_thr, as_pointcloud, mask_sky,
|
| 155 |
+
clean_depth, transparent_cams, cam_size, TSDF_thresh],
|
| 156 |
+
outputs=outmodel)
|
| 157 |
+
mask_sky.change(fn=model_from_scene_fun,
|
| 158 |
+
inputs=[scene, min_conf_thr, as_pointcloud, mask_sky,
|
| 159 |
+
clean_depth, transparent_cams, cam_size, TSDF_thresh],
|
| 160 |
+
outputs=outmodel)
|
| 161 |
+
clean_depth.change(fn=model_from_scene_fun,
|
| 162 |
+
inputs=[scene, min_conf_thr, as_pointcloud, mask_sky,
|
| 163 |
+
clean_depth, transparent_cams, cam_size, TSDF_thresh],
|
| 164 |
+
outputs=outmodel)
|
| 165 |
+
transparent_cams.change(model_from_scene_fun,
|
| 166 |
+
inputs=[scene, min_conf_thr, as_pointcloud, mask_sky,
|
| 167 |
+
clean_depth, transparent_cams, cam_size, TSDF_thresh],
|
| 168 |
+
outputs=outmodel)
|
| 169 |
+
demo.launch(share=None, server_name=None, server_port=None)
|