Spaces:
Sleeping
Sleeping
update interface
Browse files
app.py
CHANGED
|
@@ -282,7 +282,6 @@ def fuse_captions(cap1: str, cap2: str, qwen_tok, qwen_mod) -> str:
|
|
| 282 |
if fused.lower().startswith(prefix.lower()):
|
| 283 |
fused = fused[len(prefix):].strip()
|
| 284 |
|
| 285 |
-
# Safety — trim to last complete sentence if still cut off
|
| 286 |
if fused and not fused.endswith((".", "!", "?")):
|
| 287 |
last_stop = max(
|
| 288 |
fused.rfind("."),
|
|
@@ -477,17 +476,23 @@ if uploaded_file is not None:
|
|
| 477 |
with col_run:
|
| 478 |
if st.button("Generate Caption", type="primary", use_container_width=True):
|
| 479 |
|
| 480 |
-
|
|
|
|
|
|
|
|
|
|
| 481 |
(
|
| 482 |
florence_proc, florence_mod,
|
| 483 |
blip_proc, blip_itm,
|
| 484 |
qwen_tok, qwen_mod
|
| 485 |
) = load_local_models()
|
| 486 |
|
|
|
|
|
|
|
| 487 |
progress = st.progress(0)
|
| 488 |
status = st.empty()
|
| 489 |
|
| 490 |
status.info("Step 1/6: Generating captions with Florence-2-Large...")
|
|
|
|
| 491 |
captions = generate_captions_florence(
|
| 492 |
input_image, florence_proc, florence_mod
|
| 493 |
)
|
|
@@ -536,6 +541,7 @@ if uploaded_file is not None:
|
|
| 536 |
st.info(f"2. {best_2}")
|
| 537 |
|
| 538 |
status.info("Step 6/6: Fusing captions with Qwen2.5-1.5B...")
|
|
|
|
| 539 |
final = fuse_captions(best_1, best_2, qwen_tok, qwen_mod)
|
| 540 |
progress.progress(100)
|
| 541 |
status.success("Pipeline complete!")
|
|
|
|
| 282 |
if fused.lower().startswith(prefix.lower()):
|
| 283 |
fused = fused[len(prefix):].strip()
|
| 284 |
|
|
|
|
| 285 |
if fused and not fused.endswith((".", "!", "?")):
|
| 286 |
last_stop = max(
|
| 287 |
fused.rfind("."),
|
|
|
|
| 476 |
with col_run:
|
| 477 |
if st.button("Generate Caption", type="primary", use_container_width=True):
|
| 478 |
|
| 479 |
+
st.info(" First time? Models are downloading and loading — this takes 3 to 4 minutes only once. After that, every run will be much faster!")
|
| 480 |
+
st.warning(" Please do not refresh or close the page while models are loading.")
|
| 481 |
+
|
| 482 |
+
with st.spinner("⏳ Loading Florence-2, BLIP and Qwen models... Please wait patiently!"):
|
| 483 |
(
|
| 484 |
florence_proc, florence_mod,
|
| 485 |
blip_proc, blip_itm,
|
| 486 |
qwen_tok, qwen_mod
|
| 487 |
) = load_local_models()
|
| 488 |
|
| 489 |
+
st.success(" All models loaded successfully! Starting the pipeline now...")
|
| 490 |
+
|
| 491 |
progress = st.progress(0)
|
| 492 |
status = st.empty()
|
| 493 |
|
| 494 |
status.info("Step 1/6: Generating captions with Florence-2-Large...")
|
| 495 |
+
st.info(" Florence-2 is reading your image carefully — generating 5 diverse captions. This takes about 20–30 seconds, please hold on!")
|
| 496 |
captions = generate_captions_florence(
|
| 497 |
input_image, florence_proc, florence_mod
|
| 498 |
)
|
|
|
|
| 541 |
st.info(f"2. {best_2}")
|
| 542 |
|
| 543 |
status.info("Step 6/6: Fusing captions with Qwen2.5-1.5B...")
|
| 544 |
+
st.success(" Best captions selected! Now crafting your final detailed caption — please hold on for about 30–60 seconds, it is definitely on the way!")
|
| 545 |
final = fuse_captions(best_1, best_2, qwen_tok, qwen_mod)
|
| 546 |
progress.progress(100)
|
| 547 |
status.success("Pipeline complete!")
|