add some basic read me
Browse files
app.py
CHANGED
@@ -225,6 +225,15 @@ examples = [
|
|
225 |
|
226 |
|
227 |
with gr.Blocks() as demo:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
228 |
with gr.Row():
|
229 |
for i in range(max_tabs):
|
230 |
with gr.Tab(f"Input {i+1}"):
|
@@ -246,7 +255,8 @@ with gr.Blocks() as demo:
|
|
246 |
with gr.Row():
|
247 |
with gr.Accordion(f"Avergage embeddings in base 64", open=False):
|
248 |
average_embedding_base64 = gr.Textbox(show_label=False)
|
249 |
-
|
|
|
250 |
with gr.Row():
|
251 |
with gr.Column(scale=1, min_width=200):
|
252 |
scale = gr.Slider(0, 25, value=3, step=1, label="Guidance scale")
|
@@ -256,8 +266,6 @@ with gr.Blocks() as demo:
|
|
256 |
steps = gr.Slider(5, 50, value=25, step=5, label="Steps")
|
257 |
with gr.Column(scale=1, min_width=200):
|
258 |
seed = gr.Number(None, label="Seed", precision=0)
|
259 |
-
with gr.Row():
|
260 |
-
submit = gr.Button("Submit")
|
261 |
with gr.Row():
|
262 |
output = gr.Gallery(label="Generated variations")
|
263 |
|
@@ -277,6 +285,26 @@ with gr.Blocks() as demo:
|
|
277 |
submit.click(main, inputs= [average_embedding_base64, scale, n_samples, steps, seed], outputs=output)
|
278 |
output.style(grid=2)
|
279 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
280 |
|
281 |
|
282 |
if __name__ == "__main__":
|
|
|
225 |
|
226 |
|
227 |
with gr.Blocks() as demo:
|
228 |
+
with gr.Row():
|
229 |
+
gr.Markdown(
|
230 |
+
"""# Soho-Clip
|
231 |
+
|
232 |
+
A tool for exploring CLIP embedding spaces.
|
233 |
+
My interest is to use CLIP for image/video understanding (see [CLIP_visual-spatial-reasoning](https://github.com/Sohojoe/CLIP_visual-spatial-reasoning).)
|
234 |
+
|
235 |
+
Try it out by uploading a few images/add text prompts and generate images of the average of their embeddings
|
236 |
+
""")
|
237 |
with gr.Row():
|
238 |
for i in range(max_tabs):
|
239 |
with gr.Tab(f"Input {i+1}"):
|
|
|
255 |
with gr.Row():
|
256 |
with gr.Accordion(f"Avergage embeddings in base 64", open=False):
|
257 |
average_embedding_base64 = gr.Textbox(show_label=False)
|
258 |
+
with gr.Row():
|
259 |
+
submit = gr.Button("Generate images")
|
260 |
with gr.Row():
|
261 |
with gr.Column(scale=1, min_width=200):
|
262 |
scale = gr.Slider(0, 25, value=3, step=1, label="Guidance scale")
|
|
|
266 |
steps = gr.Slider(5, 50, value=25, step=5, label="Steps")
|
267 |
with gr.Column(scale=1, min_width=200):
|
268 |
seed = gr.Number(None, label="Seed", precision=0)
|
|
|
|
|
269 |
with gr.Row():
|
270 |
output = gr.Gallery(label="Generated variations")
|
271 |
|
|
|
285 |
submit.click(main, inputs= [average_embedding_base64, scale, n_samples, steps, seed], outputs=output)
|
286 |
output.style(grid=2)
|
287 |
|
288 |
+
with gr.Row():
|
289 |
+
gr.Markdown(
|
290 |
+
"""### Initial Features
|
291 |
+
|
292 |
+
- Combine up to 10 Images and/or text inputs to create an average embedding space.
|
293 |
+
- View embedding spaces as graph
|
294 |
+
- Generate a new image based on the average embedding space
|
295 |
+
|
296 |
+
### Known limitations
|
297 |
+
|
298 |
+
- Text input is a little off (requires fine tuning and I'm having issues with that at the moment)
|
299 |
+
- It can only generate a single image at a time
|
300 |
+
- Not easy to use the sample images
|
301 |
+
|
302 |
+
### Acknowledgements
|
303 |
+
|
304 |
+
- I heavily build on Justin Pinkney's Experiments in Image Variation (see https://www.justinpinkney.com/image-variation-experiments). Please credit them if you use this work.
|
305 |
+
|
306 |
+
""")
|
307 |
+
|
308 |
|
309 |
|
310 |
if __name__ == "__main__":
|