import gradio as gr
import datetime
import Implementation as imp
import time
import os
# ========== Fonctions Backend ==========
def detect_hate_speech(video_path, mode):
if mode == "Low 🌱":
Co2_release = "low"
elif mode == "Medium ♨️":
Co2_release = "medium"
else:
Co2_release = "high"
hate_speech_time, C02_emissions = imp.detectHateSpeechSmartFilter(video_path, Co2_release)
return hate_speech_time, C02_emissions
def detect_hate_speech_audio(audio_path, mode):
if mode == "Low 🌱":
Co2_release = "low"
elif mode == "Medium ♨️":
Co2_release = "medium"
else:
Co2_release = "high"
hate_speech_time, C02_emissions = imp.Detect_hate_speech_emo_hate_bert(audio_path, Co2_release)
return hate_speech_time, C02_emissions
def convertir_timestamp_en_secondes(timestamp):
h, m, s = map(int, timestamp.split(":"))
return h * 3600 + m * 60 + s
def analyser_media(fichier, mode, is_audio=False):
if not fichier:
return "
❌ No files uploaded. Please add video or audio.
", ""
if is_audio:
timestamps, carbone = detect_hate_speech_audio(fichier, mode)
else:
timestamps, carbone = detect_hate_speech(fichier, mode)
liens = ""
total_seconds = 0
for idx, (start_time, end_time) in enumerate(timestamps, 1):
secondes = convertir_timestamp_en_secondes(start_time)
liens += f'🕑 Segment {idx} : {start_time} ➔ {end_time} '
duree_segment = convertir_timestamp_en_secondes(end_time) - secondes
total_seconds += duree_segment
liens += "
"
nb_segments = len(timestamps)
duree_totale = str(datetime.timedelta(seconds=total_seconds))
resume = f"🧮 Segments detected : {nb_segments} ⏳ Total Hate Speech Duration : {duree_totale} ♻️ Carbon Footprint : {carbone}
"
return liens, resume
def afficher_pipeline(show):
return gr.update(visible=show)
def show_loader():
return gr.update(visible=True), "", ""
def analyser_avec_loading(video, mode):
liens, resume = analyser_media(video, mode, is_audio=False)
return gr.update(visible=False), liens, resume
def analyser_audio_avec_loading(audio, mode):
liens, resume = analyser_media(audio, mode, is_audio=True)
return gr.update(visible=False), liens, resume
# ========== Interface Gradio ==========
with gr.Blocks(theme=gr.themes.Monochrome(), css="body {background-color: #121212; color: white;}") as demo:
# En-tête
gr.HTML("""
🎓 EPFL Project –Emotion & Eco- Aware Hate Speech Detection in Video & Audio
This project provides an intelligent and environmentally conscious platform for detecting hate speech in videos and audio. It combines the latest tools in NLP, emotion analysis and computer vision, with CO₂ tracking, to offer both performance and eco-responsibility.
Participants: Loris Alan Fabbro, Mohammed Al-Hussini, Loic Misenta
🔗 GitHub
""")
gr.Image("logo.png", width=150, show_label=False, show_download_button=False)
# Affichage du pipeline
with gr.Row():
show_pipeline = gr.Checkbox(label="👀 Show Pipeline Overview", value=False)
pipeline_image = gr.Image(
value="pipeline.png",
label="Pipeline Overview",
show_label=True,
visible=False
)
show_pipeline.change(
afficher_pipeline,
inputs=[show_pipeline],
outputs=[pipeline_image]
)
gr.Markdown("# 🎥 Hate Speech Detector in Your Videos or Audio", elem_id="titre")
with gr.Row():
video_input = gr.Video(label="Upload your video", elem_id="video-player")
with gr.Row():
audio_input = gr.Audio(label="Upload your audio", type="filepath")
with gr.Row():
mode_selection = gr.Radio(["Low 🌱", "Medium ♨️", "High Consumption ⚠️"], label="Carbon Footprint Mode")
bouton_analyse_video = gr.Button("Detect Hate Speech in Video 🔥")
bouton_analyse_audio = gr.Button("Detect Hate Speech in Audio 🎧")
with gr.Column() as resultats:
loading_gif = gr.Image(
value="loading.gif",
visible=False,
show_label=False
)
liens_resultats = gr.HTML()
resume_resultats = gr.HTML()
bouton_analyse_video.click(
fn=show_loader,
inputs=[],
outputs=[loading_gif, liens_resultats, resume_resultats],
show_progress=False
)
bouton_analyse_video.click(
fn=analyser_avec_loading,
inputs=[video_input, mode_selection],
outputs=[loading_gif, liens_resultats, resume_resultats],
show_progress=True
)
bouton_analyse_audio.click(
fn=show_loader,
inputs=[],
outputs=[loading_gif, liens_resultats, resume_resultats],
show_progress=False
)
bouton_analyse_audio.click(
fn=analyser_audio_avec_loading,
inputs=[audio_input, mode_selection],
outputs=[loading_gif, liens_resultats, resume_resultats],
show_progress=True
)
demo.launch(share=True)