Vishal Patil commited on
Update app.py
Browse files
app.py
CHANGED
|
@@ -2,22 +2,10 @@ import tempfile
|
|
| 2 |
import traceback
|
| 3 |
from pathlib import Path
|
| 4 |
import gradio as gr
|
| 5 |
-
import
|
| 6 |
-
|
| 7 |
-
# Load pipeline module explicitly to avoid circular import
|
| 8 |
-
spec = importlib.util.spec_from_file_location("speech_pipeline", "/app/pipeline.py")
|
| 9 |
-
mod = importlib.util.module_from_spec(spec)
|
| 10 |
-
sys.modules["speech_pipeline"] = mod
|
| 11 |
-
spec.loader.exec_module(mod)
|
| 12 |
-
|
| 13 |
-
get_pipeline = mod.get_pipeline
|
| 14 |
-
to_json = mod.to_json
|
| 15 |
-
to_csv = mod.to_csv
|
| 16 |
-
to_abab_text = mod.to_abab_text
|
| 17 |
|
| 18 |
UPLOAD_DIR = Path(tempfile.gettempdir()) / "speech_annotation"
|
| 19 |
UPLOAD_DIR.mkdir(parents=True, exist_ok=True)
|
| 20 |
-
|
| 21 |
_last_segments = []
|
| 22 |
|
| 23 |
def process_audio(audio_path, num_speakers):
|
|
@@ -33,20 +21,17 @@ def process_audio(audio_path, num_speakers):
|
|
| 33 |
if not segments:
|
| 34 |
return "⚠️ No speech detected.", ""
|
| 35 |
_last_segments = segments
|
| 36 |
-
abab = to_abab_text(segments)
|
| 37 |
unique = len(set(s.speaker for s in segments))
|
| 38 |
-
return f"✅ Done — {len(segments)} segments, {unique} speaker(s) detected",
|
| 39 |
|
| 40 |
def export_json():
|
| 41 |
-
if not _last_segments:
|
| 42 |
-
return None
|
| 43 |
out = str(UPLOAD_DIR / "annotation.json")
|
| 44 |
to_json(_last_segments, out)
|
| 45 |
return out
|
| 46 |
|
| 47 |
def export_csv():
|
| 48 |
-
if not _last_segments:
|
| 49 |
-
return None
|
| 50 |
out = str(UPLOAD_DIR / "annotation.csv")
|
| 51 |
to_csv(_last_segments, out)
|
| 52 |
return out
|
|
@@ -56,21 +41,20 @@ with gr.Blocks(title="Speech Annotation Pipeline") as demo:
|
|
| 56 |
audio_input = gr.Audio(label="Upload Audio (.wav/.mp3/.flac)", type="filepath")
|
| 57 |
num_speakers = gr.Slider(minimum=0, maximum=10, step=1, value=0,
|
| 58 |
label="Number of speakers (0 = auto-detect)")
|
| 59 |
-
run_btn
|
| 60 |
-
status_box
|
| 61 |
-
abab_box
|
| 62 |
-
|
| 63 |
with gr.Row():
|
| 64 |
json_btn = gr.Button("Export JSON")
|
| 65 |
csv_btn = gr.Button("Export CSV")
|
| 66 |
json_file = gr.File(label="JSON")
|
| 67 |
csv_file = gr.File(label="CSV")
|
| 68 |
|
| 69 |
-
run_btn.click(fn=process_audio,
|
| 70 |
-
inputs=[audio_input, num_speakers],
|
| 71 |
outputs=[status_box, abab_box])
|
| 72 |
json_btn.click(fn=export_json, inputs=[], outputs=[json_file])
|
| 73 |
-
csv_btn.click(fn=export_csv,
|
| 74 |
|
| 75 |
if __name__ == "__main__":
|
| 76 |
demo.launch(server_name="0.0.0.0", server_port=7860, show_error=True)
|
|
|
|
| 2 |
import traceback
|
| 3 |
from pathlib import Path
|
| 4 |
import gradio as gr
|
| 5 |
+
from pipeline import get_pipeline, to_json, to_csv, to_abab_text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
|
| 7 |
UPLOAD_DIR = Path(tempfile.gettempdir()) / "speech_annotation"
|
| 8 |
UPLOAD_DIR.mkdir(parents=True, exist_ok=True)
|
|
|
|
| 9 |
_last_segments = []
|
| 10 |
|
| 11 |
def process_audio(audio_path, num_speakers):
|
|
|
|
| 21 |
if not segments:
|
| 22 |
return "⚠️ No speech detected.", ""
|
| 23 |
_last_segments = segments
|
|
|
|
| 24 |
unique = len(set(s.speaker for s in segments))
|
| 25 |
+
return f"✅ Done — {len(segments)} segments, {unique} speaker(s) detected", to_abab_text(segments)
|
| 26 |
|
| 27 |
def export_json():
|
| 28 |
+
if not _last_segments: return None
|
|
|
|
| 29 |
out = str(UPLOAD_DIR / "annotation.json")
|
| 30 |
to_json(_last_segments, out)
|
| 31 |
return out
|
| 32 |
|
| 33 |
def export_csv():
|
| 34 |
+
if not _last_segments: return None
|
|
|
|
| 35 |
out = str(UPLOAD_DIR / "annotation.csv")
|
| 36 |
to_csv(_last_segments, out)
|
| 37 |
return out
|
|
|
|
| 41 |
audio_input = gr.Audio(label="Upload Audio (.wav/.mp3/.flac)", type="filepath")
|
| 42 |
num_speakers = gr.Slider(minimum=0, maximum=10, step=1, value=0,
|
| 43 |
label="Number of speakers (0 = auto-detect)")
|
| 44 |
+
run_btn = gr.Button("▶ Run Annotation", variant="primary")
|
| 45 |
+
status_box = gr.Textbox(label="Status", value="Ready.", interactive=False)
|
| 46 |
+
abab_box = gr.Textbox(label="Transcript (ABAB)", lines=15, value="",
|
| 47 |
+
placeholder="Annotations will appear here...")
|
| 48 |
with gr.Row():
|
| 49 |
json_btn = gr.Button("Export JSON")
|
| 50 |
csv_btn = gr.Button("Export CSV")
|
| 51 |
json_file = gr.File(label="JSON")
|
| 52 |
csv_file = gr.File(label="CSV")
|
| 53 |
|
| 54 |
+
run_btn.click(fn=process_audio, inputs=[audio_input, num_speakers],
|
|
|
|
| 55 |
outputs=[status_box, abab_box])
|
| 56 |
json_btn.click(fn=export_json, inputs=[], outputs=[json_file])
|
| 57 |
+
csv_btn.click(fn=export_csv, inputs=[], outputs=[csv_file])
|
| 58 |
|
| 59 |
if __name__ == "__main__":
|
| 60 |
demo.launch(server_name="0.0.0.0", server_port=7860, show_error=True)
|