Spaces:
Running
on
Zero
Running
on
Zero
update cluster plot
Browse files
app.py
CHANGED
@@ -1,10 +1,13 @@
|
|
1 |
# Author: Huzheng Yang
|
2 |
# %%
|
3 |
import copy
|
|
|
4 |
from io import BytesIO
|
5 |
import os
|
6 |
|
|
|
7 |
from matplotlib import pyplot as plt
|
|
|
8 |
USE_HUGGINGFACE_ZEROGPU = os.getenv("USE_HUGGINGFACE_ZEROGPU", "False").lower() in ["true", "1", "yes"]
|
9 |
DOWNLOAD_ALL_MODELS_DATASETS = os.getenv("DOWNLOAD_ALL_MODELS_DATASETS", "False").lower() in ["true", "1", "yes"]
|
10 |
|
@@ -219,17 +222,111 @@ def run_alignedthreemodelattnnodes(images, model, batch_size=16):
|
|
219 |
return outputs
|
220 |
|
221 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
222 |
def ncut_run(
|
223 |
model,
|
224 |
images,
|
225 |
-
model_name="
|
226 |
-
layer
|
227 |
num_eig=100,
|
228 |
node_type="block",
|
229 |
-
affinity_focal_gamma=0.
|
230 |
num_sample_ncut=10000,
|
231 |
knn_ncut=10,
|
232 |
-
embedding_method="
|
233 |
embedding_metric='euclidean',
|
234 |
num_sample_tsne=1000,
|
235 |
knn_tsne=10,
|
@@ -353,8 +450,10 @@ def ncut_run(
|
|
353 |
logging_str += _logging_str
|
354 |
rgb.append(_rgb[0])
|
355 |
|
356 |
-
|
357 |
-
|
|
|
|
|
358 |
features,
|
359 |
num_eig=num_eig,
|
360 |
num_sample_ncut=num_sample_ncut,
|
@@ -384,7 +483,6 @@ def ncut_run(
|
|
384 |
pil_images.append(_im)
|
385 |
return pil_images, logging_str
|
386 |
|
387 |
-
|
388 |
|
389 |
if is_lisa == True:
|
390 |
# dirty patch for the LISA model
|
@@ -396,16 +494,26 @@ def ncut_run(
|
|
396 |
|
397 |
rgb = dont_use_too_much_green(rgb)
|
398 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
399 |
|
400 |
if video_output:
|
401 |
video_path = get_random_path()
|
402 |
video_cache.add_video(video_path)
|
403 |
pil_images_to_video(to_pil_images(rgb), video_path)
|
404 |
return video_path, logging_str
|
405 |
-
|
406 |
-
|
|
|
|
|
407 |
|
408 |
def _ncut_run(*args, **kwargs):
|
|
|
409 |
try:
|
410 |
if torch.cuda.is_available():
|
411 |
torch.cuda.empty_cache()
|
@@ -414,15 +522,17 @@ def _ncut_run(*args, **kwargs):
|
|
414 |
|
415 |
if torch.cuda.is_available():
|
416 |
torch.cuda.empty_cache()
|
417 |
-
|
|
|
418 |
return ret
|
419 |
except Exception as e:
|
420 |
gr.Error(str(e))
|
421 |
if torch.cuda.is_available():
|
422 |
torch.cuda.empty_cache()
|
423 |
-
return
|
424 |
|
425 |
# ret = ncut_run(*args, **kwargs)
|
|
|
426 |
# return ret
|
427 |
|
428 |
if USE_HUGGINGFACE_ZEROGPU:
|
@@ -488,6 +598,16 @@ def transform_image(image, resolution=(1024, 1024), stablediffusion=False):
|
|
488 |
image = image * 2 - 1
|
489 |
return image
|
490 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
491 |
def plot_one_image_36_grid(original_image, tsne_rgb_images):
|
492 |
mean = [0.485, 0.456, 0.406]
|
493 |
std = [0.229, 0.224, 0.225]
|
@@ -583,8 +703,8 @@ promptable_segmentation_models = ["LISA(xinlai/LISA-7B-v1)"]
|
|
583 |
|
584 |
def run_fn(
|
585 |
images,
|
586 |
-
model_name="
|
587 |
-
layer
|
588 |
num_eig=100,
|
589 |
node_type="block",
|
590 |
positive_prompt="",
|
@@ -593,15 +713,15 @@ def run_fn(
|
|
593 |
lisa_prompt1="",
|
594 |
lisa_prompt2="",
|
595 |
lisa_prompt3="",
|
596 |
-
affinity_focal_gamma=0.
|
597 |
num_sample_ncut=10000,
|
598 |
knn_ncut=10,
|
599 |
-
embedding_method="
|
600 |
embedding_metric='euclidean',
|
601 |
-
num_sample_tsne=
|
602 |
knn_tsne=10,
|
603 |
-
perplexity=
|
604 |
-
n_neighbors=
|
605 |
min_dist=0.1,
|
606 |
sampling_method="fps",
|
607 |
old_school_ncut=False,
|
@@ -613,11 +733,12 @@ def run_fn(
|
|
613 |
recursion_l1_gamma=0.5,
|
614 |
recursion_l2_gamma=0.5,
|
615 |
recursion_l3_gamma=0.5,
|
|
|
616 |
):
|
617 |
|
618 |
if images is None:
|
619 |
gr.Warning("No images selected.")
|
620 |
-
return
|
621 |
|
622 |
video_output = False
|
623 |
if isinstance(images, str):
|
@@ -733,6 +854,7 @@ def run_fn(
|
|
733 |
"lisa_prompt2": lisa_prompt2,
|
734 |
"lisa_prompt3": lisa_prompt3,
|
735 |
"is_lisa": is_lisa,
|
|
|
736 |
}
|
737 |
# print(kwargs)
|
738 |
|
@@ -1042,9 +1164,11 @@ with demo:
|
|
1042 |
with gr.Column(scale=5, min_width=200):
|
1043 |
input_gallery, submit_button, clear_images_button = make_input_images_section()
|
1044 |
dataset_dropdown, num_images_slider, random_seed_slider, load_images_button = make_dataset_images_section()
|
|
|
1045 |
|
1046 |
with gr.Column(scale=5, min_width=200):
|
1047 |
output_gallery = make_output_images_section()
|
|
|
1048 |
[
|
1049 |
model_dropdown, layer_slider, node_type_dropdown, num_eig_slider,
|
1050 |
affinity_focal_gamma_slider, num_sample_ncut_slider, knn_ncut_slider,
|
@@ -1052,16 +1176,15 @@ with demo:
|
|
1052 |
perplexity_slider, n_neighbors_slider, min_dist_slider,
|
1053 |
sampling_method_dropdown, positive_prompt, negative_prompt
|
1054 |
] = make_parameters_section()
|
1055 |
-
|
1056 |
-
logging_text = gr.Textbox("Logging information", label="Logging", elem_id="logging", type="text", placeholder="Logging information")
|
1057 |
|
1058 |
-
clear_images_button.click(lambda x: ([], []), outputs=[input_gallery, output_gallery])
|
1059 |
|
1060 |
false_placeholder = gr.Checkbox(label="False", value=False, elem_id="false_placeholder", visible=False)
|
1061 |
no_prompt = gr.Textbox("", label="", elem_id="empty_placeholder", type="text", placeholder="", visible=False)
|
1062 |
|
1063 |
submit_button.click(
|
1064 |
-
run_fn,
|
1065 |
inputs=[
|
1066 |
input_gallery, model_dropdown, layer_slider, num_eig_slider, node_type_dropdown,
|
1067 |
positive_prompt, negative_prompt,
|
@@ -1070,7 +1193,7 @@ with demo:
|
|
1070 |
embedding_method_dropdown, embedding_metric_dropdown, num_sample_tsne_slider, knn_tsne_slider,
|
1071 |
perplexity_slider, n_neighbors_slider, min_dist_slider, sampling_method_dropdown
|
1072 |
],
|
1073 |
-
outputs=[output_gallery, logging_text],
|
1074 |
api_name="API_AlignedCut"
|
1075 |
)
|
1076 |
|
@@ -1201,7 +1324,7 @@ with demo:
|
|
1201 |
no_prompt = gr.Textbox("", label="", elem_id="empty_placeholder", type="text", placeholder="", visible=False)
|
1202 |
|
1203 |
submit_button.click(
|
1204 |
-
run_fn,
|
1205 |
inputs=[
|
1206 |
input_gallery, model_dropdown, layer_slider, num_eig_slider, node_type_dropdown,
|
1207 |
positive_prompt, negative_prompt,
|
@@ -1305,7 +1428,7 @@ with demo:
|
|
1305 |
galleries = [l1_gallery, l2_gallery, l3_gallery]
|
1306 |
true_placeholder = gr.Checkbox(label="True placeholder", value=True, elem_id="true_placeholder", visible=False)
|
1307 |
submit_button.click(
|
1308 |
-
run_fn,
|
1309 |
inputs=[
|
1310 |
input_gallery, model_dropdown, layer_slider, num_eig_slider, node_type_dropdown,
|
1311 |
positive_prompt, negative_prompt,
|
@@ -1465,6 +1588,7 @@ with demo:
|
|
1465 |
gr.Markdown("**This demo is for the Python package `ncut-pytorch`, please visit the [Documentation](https://ncut-pytorch.readthedocs.io/)**")
|
1466 |
gr.Markdown("**All the models and functions used for this demo are in the Python package `ncut-pytorch`**")
|
1467 |
gr.Markdown("---")
|
|
|
1468 |
gr.Markdown("**Normalized Cuts**, aka. spectral clustering, is a graphical method to analyze data grouping in the affinity eigenvector space. It has been widely used for unsupervised segmentation in the 2000s.")
|
1469 |
gr.Markdown("*Normalized Cuts and Image Segmentation, Jianbo Shi and Jitendra Malik, 2000*")
|
1470 |
gr.Markdown("---")
|
@@ -1473,7 +1597,9 @@ with demo:
|
|
1473 |
gr.Markdown("- **spectral-tSNE** visualization, a new method to visualize the high-dimensional eigenvector space with 3D RGB cube. Color is aligned across images, color infers distance in representation.")
|
1474 |
gr.Markdown("*paper in prep, Yang 2024*")
|
1475 |
gr.Markdown("*AlignedCut: Visual Concepts Discovery on Brain-Guided Universal Feature Space, Huzheng Yang, James Gee\*, and Jianbo Shi\*, 2024*")
|
1476 |
-
|
|
|
|
|
1477 |
|
1478 |
|
1479 |
with gr.Row():
|
@@ -1497,4 +1623,10 @@ if DOWNLOAD_ALL_MODELS_DATASETS:
|
|
1497 |
demo.launch(share=True)
|
1498 |
|
1499 |
|
1500 |
-
# %%
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
# Author: Huzheng Yang
|
2 |
# %%
|
3 |
import copy
|
4 |
+
from functools import partial
|
5 |
from io import BytesIO
|
6 |
import os
|
7 |
|
8 |
+
from einops import rearrange
|
9 |
from matplotlib import pyplot as plt
|
10 |
+
import matplotlib
|
11 |
USE_HUGGINGFACE_ZEROGPU = os.getenv("USE_HUGGINGFACE_ZEROGPU", "False").lower() in ["true", "1", "yes"]
|
12 |
DOWNLOAD_ALL_MODELS_DATASETS = os.getenv("DOWNLOAD_ALL_MODELS_DATASETS", "False").lower() in ["true", "1", "yes"]
|
13 |
|
|
|
222 |
return outputs
|
223 |
|
224 |
|
225 |
+
def _reds_colormap(image):
|
226 |
+
# normed_data = image / image.max() # Normalize to [0, 1]
|
227 |
+
normed_data = image
|
228 |
+
colormap = matplotlib.colormaps['inferno'] # Get the Reds colormap
|
229 |
+
colored_image = colormap(normed_data) # Apply colormap
|
230 |
+
return (colored_image[..., :3] * 255).astype(np.uint8) # Convert to RGB
|
231 |
+
|
232 |
+
# heatmap images
|
233 |
+
def apply_reds_colormap(images, size):
|
234 |
+
# for i_image in range(images.shape[0]):
|
235 |
+
# images[i_image] -= images[i_image].min()
|
236 |
+
# images[i_image] /= images[i_image].max()
|
237 |
+
# normed_data = [_reds_colormap(images[i]) for i in range(images.shape[0])]
|
238 |
+
# normed_data = np.stack(normed_data)
|
239 |
+
normed_data = _reds_colormap(images)
|
240 |
+
normed_data = torch.tensor(normed_data).float()
|
241 |
+
normed_data = rearrange(normed_data, "b h w c -> b c h w")
|
242 |
+
normed_data = torch.nn.functional.interpolate(normed_data, size=size, mode="nearest")
|
243 |
+
normed_data = rearrange(normed_data, "b c h w -> b h w c")
|
244 |
+
normed_data = normed_data.cpu().numpy().astype(np.uint8)
|
245 |
+
return normed_data
|
246 |
+
|
247 |
+
# Blend heatmap with the original image
|
248 |
+
def blend_image_with_heatmap(image, heatmap, opacity1=0.5, opacity2=0.5):
|
249 |
+
blended = (1 - opacity1) * image + opacity2 * heatmap
|
250 |
+
return blended.astype(np.uint8)
|
251 |
+
|
252 |
+
def make_cluster_plot(eigvecs, images, h=64, w=64):
|
253 |
+
from ncut_pytorch.ncut_pytorch import farthest_point_sampling
|
254 |
+
magnitude = torch.norm(eigvecs, dim=-1)
|
255 |
+
p = 0.5
|
256 |
+
top_p_idx = magnitude.argsort(descending=True)[:int(p * magnitude.shape[0])]
|
257 |
+
num_samples = 50
|
258 |
+
fps_idx = farthest_point_sampling(eigvecs[top_p_idx], num_samples)
|
259 |
+
fps_idx = top_p_idx[fps_idx]
|
260 |
+
|
261 |
+
# downsample to 256x256
|
262 |
+
images = F.interpolate(images, (256, 256), mode="bilinear")
|
263 |
+
images = images.cpu().numpy()
|
264 |
+
images = images.transpose(0, 2, 3, 1)
|
265 |
+
images = images * 255
|
266 |
+
images = images.astype(np.uint8)
|
267 |
+
|
268 |
+
|
269 |
+
# sort the fps_idx by the mean of the heatmap
|
270 |
+
fps_heatmaps = {}
|
271 |
+
sort_values = []
|
272 |
+
for _, idx in enumerate(fps_idx):
|
273 |
+
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
274 |
+
eigvecs = eigvecs.to(device)
|
275 |
+
heatmap = F.cosine_similarity(eigvecs, eigvecs[idx][None], dim=-1)
|
276 |
+
heatmap = heatmap.reshape(-1, h, w)
|
277 |
+
mask = (heatmap > 0.5).float()
|
278 |
+
sort_values.append(mask.mean().item())
|
279 |
+
fps_heatmaps[idx.item()] = heatmap.cpu()
|
280 |
+
|
281 |
+
fig_images = []
|
282 |
+
i_cluster = 0
|
283 |
+
for i_fig in range(10):
|
284 |
+
fig, axs = plt.subplots(3, 5, figsize=(15, 9))
|
285 |
+
for ax in axs.flatten():
|
286 |
+
ax.axis("off")
|
287 |
+
for j, idx in enumerate(fps_idx[i_fig*5:i_fig*5+5]):
|
288 |
+
heatmap = fps_heatmaps[idx.item()]
|
289 |
+
mask = (heatmap > 0.1).float()
|
290 |
+
sorted_image_idxs = torch.argsort(mask.mean((1, 2)), descending=True)
|
291 |
+
size = (images.shape[1], images.shape[2])
|
292 |
+
heatmap = apply_reds_colormap(heatmap, size)
|
293 |
+
for i, image_idx in enumerate(sorted_image_idxs[:3]):
|
294 |
+
_heatmap = blend_image_with_heatmap(images[image_idx], heatmap[image_idx])
|
295 |
+
axs[i, j].imshow(_heatmap)
|
296 |
+
if i == 0:
|
297 |
+
axs[i, j].set_title(f"cluster {i_cluster+1}", fontsize=24)
|
298 |
+
i_cluster += 1
|
299 |
+
plt.tight_layout(h_pad=0.5, w_pad=0.3)
|
300 |
+
|
301 |
+
buf = BytesIO()
|
302 |
+
plt.savefig(buf, bbox_inches='tight', dpi=72)
|
303 |
+
|
304 |
+
buf.seek(0) # Move to the start of the BytesIO buffer
|
305 |
+
img = Image.open(buf)
|
306 |
+
img = img.convert("RGB")
|
307 |
+
img = copy.deepcopy(img)
|
308 |
+
buf.close()
|
309 |
+
|
310 |
+
fig_images.append(img)
|
311 |
+
plt.close()
|
312 |
+
|
313 |
+
# plt.imshow(img)
|
314 |
+
# plt.axis("off")
|
315 |
+
# plt.show()
|
316 |
+
return fig_images
|
317 |
+
|
318 |
+
|
319 |
def ncut_run(
|
320 |
model,
|
321 |
images,
|
322 |
+
model_name="DiNO(dino_vitb8_448)",
|
323 |
+
layer=10,
|
324 |
num_eig=100,
|
325 |
node_type="block",
|
326 |
+
affinity_focal_gamma=0.5,
|
327 |
num_sample_ncut=10000,
|
328 |
knn_ncut=10,
|
329 |
+
embedding_method="tsne_3d",
|
330 |
embedding_metric='euclidean',
|
331 |
num_sample_tsne=1000,
|
332 |
knn_tsne=10,
|
|
|
450 |
logging_str += _logging_str
|
451 |
rgb.append(_rgb[0])
|
452 |
|
453 |
+
|
454 |
+
cluster_images = None
|
455 |
+
if not old_school_ncut: # ailgnedcut, joint across all images
|
456 |
+
rgb, _logging_str, eigvecs = compute_ncut(
|
457 |
features,
|
458 |
num_eig=num_eig,
|
459 |
num_sample_ncut=num_sample_ncut,
|
|
|
483 |
pil_images.append(_im)
|
484 |
return pil_images, logging_str
|
485 |
|
|
|
486 |
|
487 |
if is_lisa == True:
|
488 |
# dirty patch for the LISA model
|
|
|
494 |
|
495 |
rgb = dont_use_too_much_green(rgb)
|
496 |
|
497 |
+
if not video_output:
|
498 |
+
start = time.time()
|
499 |
+
h, w = features.shape[1], features.shape[2]
|
500 |
+
_images = reverse_transform_image(images, stablediffusion="stable" in model_name.lower())
|
501 |
+
cluster_images = make_cluster_plot(eigvecs, _images, h=h, w=w)
|
502 |
+
logging_str += f"Plot time: {time.time() - start:.2f}s\n"
|
503 |
+
|
504 |
|
505 |
if video_output:
|
506 |
video_path = get_random_path()
|
507 |
video_cache.add_video(video_path)
|
508 |
pil_images_to_video(to_pil_images(rgb), video_path)
|
509 |
return video_path, logging_str
|
510 |
+
|
511 |
+
|
512 |
+
return to_pil_images(rgb), cluster_images, logging_str
|
513 |
+
|
514 |
|
515 |
def _ncut_run(*args, **kwargs):
|
516 |
+
n_ret = kwargs.pop("n_ret", 1)
|
517 |
try:
|
518 |
if torch.cuda.is_available():
|
519 |
torch.cuda.empty_cache()
|
|
|
522 |
|
523 |
if torch.cuda.is_available():
|
524 |
torch.cuda.empty_cache()
|
525 |
+
|
526 |
+
ret = list(ret)[:n_ret] + [ret[-1]]
|
527 |
return ret
|
528 |
except Exception as e:
|
529 |
gr.Error(str(e))
|
530 |
if torch.cuda.is_available():
|
531 |
torch.cuda.empty_cache()
|
532 |
+
return *(None for _ in range(n_ret)), "Error: " + str(e)
|
533 |
|
534 |
# ret = ncut_run(*args, **kwargs)
|
535 |
+
# ret = list(ret)[:n_ret] + [ret[-1]]
|
536 |
# return ret
|
537 |
|
538 |
if USE_HUGGINGFACE_ZEROGPU:
|
|
|
598 |
image = image * 2 - 1
|
599 |
return image
|
600 |
|
601 |
+
def reverse_transform_image(image, stablediffusion=False):
|
602 |
+
if stablediffusion:
|
603 |
+
image = (image + 1) / 2
|
604 |
+
else:
|
605 |
+
mean = [0.485, 0.456, 0.406]
|
606 |
+
std = [0.229, 0.224, 0.225]
|
607 |
+
image = image * torch.tensor(std).view(3, 1, 1) + torch.tensor(mean).view(3, 1, 1)
|
608 |
+
image = torch.clamp(image, 0, 1)
|
609 |
+
return image
|
610 |
+
|
611 |
def plot_one_image_36_grid(original_image, tsne_rgb_images):
|
612 |
mean = [0.485, 0.456, 0.406]
|
613 |
std = [0.229, 0.224, 0.225]
|
|
|
703 |
|
704 |
def run_fn(
|
705 |
images,
|
706 |
+
model_name="DiNO(dino_vitb8_448)",
|
707 |
+
layer=10,
|
708 |
num_eig=100,
|
709 |
node_type="block",
|
710 |
positive_prompt="",
|
|
|
713 |
lisa_prompt1="",
|
714 |
lisa_prompt2="",
|
715 |
lisa_prompt3="",
|
716 |
+
affinity_focal_gamma=0.5,
|
717 |
num_sample_ncut=10000,
|
718 |
knn_ncut=10,
|
719 |
+
embedding_method="tsne_3d",
|
720 |
embedding_metric='euclidean',
|
721 |
+
num_sample_tsne=300,
|
722 |
knn_tsne=10,
|
723 |
+
perplexity=150,
|
724 |
+
n_neighbors=150,
|
725 |
min_dist=0.1,
|
726 |
sampling_method="fps",
|
727 |
old_school_ncut=False,
|
|
|
733 |
recursion_l1_gamma=0.5,
|
734 |
recursion_l2_gamma=0.5,
|
735 |
recursion_l3_gamma=0.5,
|
736 |
+
n_ret=1,
|
737 |
):
|
738 |
|
739 |
if images is None:
|
740 |
gr.Warning("No images selected.")
|
741 |
+
return *(None for _ in range(n_ret)), "No images selected."
|
742 |
|
743 |
video_output = False
|
744 |
if isinstance(images, str):
|
|
|
854 |
"lisa_prompt2": lisa_prompt2,
|
855 |
"lisa_prompt3": lisa_prompt3,
|
856 |
"is_lisa": is_lisa,
|
857 |
+
"n_ret": n_ret,
|
858 |
}
|
859 |
# print(kwargs)
|
860 |
|
|
|
1164 |
with gr.Column(scale=5, min_width=200):
|
1165 |
input_gallery, submit_button, clear_images_button = make_input_images_section()
|
1166 |
dataset_dropdown, num_images_slider, random_seed_slider, load_images_button = make_dataset_images_section()
|
1167 |
+
logging_text = gr.Textbox("Logging information", label="Logging", elem_id="logging", type="text", placeholder="Logging information")
|
1168 |
|
1169 |
with gr.Column(scale=5, min_width=200):
|
1170 |
output_gallery = make_output_images_section()
|
1171 |
+
cluster_gallery = gr.Gallery(value=[], label="Clusters", show_label=False, elem_id="clusters", columns=[2], rows=[1], object_fit="contain", height="auto", show_share_button=True, preview=True)
|
1172 |
[
|
1173 |
model_dropdown, layer_slider, node_type_dropdown, num_eig_slider,
|
1174 |
affinity_focal_gamma_slider, num_sample_ncut_slider, knn_ncut_slider,
|
|
|
1176 |
perplexity_slider, n_neighbors_slider, min_dist_slider,
|
1177 |
sampling_method_dropdown, positive_prompt, negative_prompt
|
1178 |
] = make_parameters_section()
|
1179 |
+
num_eig_slider.value = 30
|
|
|
1180 |
|
1181 |
+
clear_images_button.click(lambda x: ([], [], []), outputs=[input_gallery, output_gallery, cluster_gallery])
|
1182 |
|
1183 |
false_placeholder = gr.Checkbox(label="False", value=False, elem_id="false_placeholder", visible=False)
|
1184 |
no_prompt = gr.Textbox("", label="", elem_id="empty_placeholder", type="text", placeholder="", visible=False)
|
1185 |
|
1186 |
submit_button.click(
|
1187 |
+
partial(run_fn, n_ret=2),
|
1188 |
inputs=[
|
1189 |
input_gallery, model_dropdown, layer_slider, num_eig_slider, node_type_dropdown,
|
1190 |
positive_prompt, negative_prompt,
|
|
|
1193 |
embedding_method_dropdown, embedding_metric_dropdown, num_sample_tsne_slider, knn_tsne_slider,
|
1194 |
perplexity_slider, n_neighbors_slider, min_dist_slider, sampling_method_dropdown
|
1195 |
],
|
1196 |
+
outputs=[output_gallery, cluster_gallery, logging_text],
|
1197 |
api_name="API_AlignedCut"
|
1198 |
)
|
1199 |
|
|
|
1324 |
no_prompt = gr.Textbox("", label="", elem_id="empty_placeholder", type="text", placeholder="", visible=False)
|
1325 |
|
1326 |
submit_button.click(
|
1327 |
+
partial(run_fn, n_ret=3),
|
1328 |
inputs=[
|
1329 |
input_gallery, model_dropdown, layer_slider, num_eig_slider, node_type_dropdown,
|
1330 |
positive_prompt, negative_prompt,
|
|
|
1428 |
galleries = [l1_gallery, l2_gallery, l3_gallery]
|
1429 |
true_placeholder = gr.Checkbox(label="True placeholder", value=True, elem_id="true_placeholder", visible=False)
|
1430 |
submit_button.click(
|
1431 |
+
partial(run_fn, n_ret=len(galleries)),
|
1432 |
inputs=[
|
1433 |
input_gallery, model_dropdown, layer_slider, num_eig_slider, node_type_dropdown,
|
1434 |
positive_prompt, negative_prompt,
|
|
|
1588 |
gr.Markdown("**This demo is for the Python package `ncut-pytorch`, please visit the [Documentation](https://ncut-pytorch.readthedocs.io/)**")
|
1589 |
gr.Markdown("**All the models and functions used for this demo are in the Python package `ncut-pytorch`**")
|
1590 |
gr.Markdown("---")
|
1591 |
+
gr.Markdown("---")
|
1592 |
gr.Markdown("**Normalized Cuts**, aka. spectral clustering, is a graphical method to analyze data grouping in the affinity eigenvector space. It has been widely used for unsupervised segmentation in the 2000s.")
|
1593 |
gr.Markdown("*Normalized Cuts and Image Segmentation, Jianbo Shi and Jitendra Malik, 2000*")
|
1594 |
gr.Markdown("---")
|
|
|
1597 |
gr.Markdown("- **spectral-tSNE** visualization, a new method to visualize the high-dimensional eigenvector space with 3D RGB cube. Color is aligned across images, color infers distance in representation.")
|
1598 |
gr.Markdown("*paper in prep, Yang 2024*")
|
1599 |
gr.Markdown("*AlignedCut: Visual Concepts Discovery on Brain-Guided Universal Feature Space, Huzheng Yang, James Gee\*, and Jianbo Shi\*, 2024*")
|
1600 |
+
gr.Markdown("---")
|
1601 |
+
gr.Markdown("---")
|
1602 |
+
gr.Markdown('<p style="text-align: center;">We thank the HuggingFace team for hosting this demo.</p>')
|
1603 |
|
1604 |
|
1605 |
with gr.Row():
|
|
|
1623 |
demo.launch(share=True)
|
1624 |
|
1625 |
|
1626 |
+
# # %%
|
1627 |
+
# # debug
|
1628 |
+
# # change working directory to "/"
|
1629 |
+
# os.chdir("/")
|
1630 |
+
# images = [(Image.open(image), None) for image in default_images]
|
1631 |
+
# ret = run_fn(images, num_eig=30)
|
1632 |
+
# # %%
|