Aleks Zhuravlev
commited on
Commit
·
6d91b68
1
Parent(s):
58ef339
Add application file
Browse files- app.py +388 -0
- requirements.txt +12 -0
app.py
ADDED
|
@@ -0,0 +1,388 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# app.py
|
| 2 |
+
import gradio as gr
|
| 3 |
+
import torch
|
| 4 |
+
import torchaudio
|
| 5 |
+
from transformers import (
|
| 6 |
+
pipeline, AutoProcessor, AutoModelForSpeechSeq2Seq,
|
| 7 |
+
AutoImageProcessor, AutoModelForObjectDetection,
|
| 8 |
+
BlipForQuestionAnswering, BlipProcessor, CLIPModel, CLIPProcessor,
|
| 9 |
+
VitsModel, AutoTokenizer
|
| 10 |
+
)
|
| 11 |
+
from PIL import Image, ImageDraw
|
| 12 |
+
import requests
|
| 13 |
+
import numpy as np
|
| 14 |
+
import soundfile as sf
|
| 15 |
+
from gtts import gTTS
|
| 16 |
+
import tempfile
|
| 17 |
+
import os
|
| 18 |
+
from sentence_transformers import SentenceTransformer
|
| 19 |
+
|
| 20 |
+
# Инициализация моделей (ленивая загрузка)
|
| 21 |
+
models = {}
|
| 22 |
+
|
| 23 |
+
def load_audio_model(model_name):
|
| 24 |
+
if model_name not in models:
|
| 25 |
+
if model_name == "whisper":
|
| 26 |
+
models[model_name] = pipeline(
|
| 27 |
+
"automatic-speech-recognition",
|
| 28 |
+
model="openai/whisper-small"
|
| 29 |
+
)
|
| 30 |
+
elif model_name == "wav2vec2":
|
| 31 |
+
models[model_name] = pipeline(
|
| 32 |
+
"automatic-speech-recognition",
|
| 33 |
+
model="bond005/wav2vec2-large-ru-golos"
|
| 34 |
+
)
|
| 35 |
+
elif model_name == "audio_classifier":
|
| 36 |
+
models[model_name] = pipeline(
|
| 37 |
+
"audio-classification",
|
| 38 |
+
model="MIT/ast-finetuned-audioset-10-10-0.4593"
|
| 39 |
+
)
|
| 40 |
+
elif model_name == "emotion_classifier":
|
| 41 |
+
models[model_name] = pipeline(
|
| 42 |
+
"audio-classification",
|
| 43 |
+
model="superb/hubert-large-superb-er"
|
| 44 |
+
)
|
| 45 |
+
return models[model_name]
|
| 46 |
+
|
| 47 |
+
def load_image_model(model_name):
|
| 48 |
+
if model_name not in models:
|
| 49 |
+
if model_name == "object_detection":
|
| 50 |
+
models[model_name] = pipeline("object-detection", model="facebook/detr-resnet-50")
|
| 51 |
+
elif model_name == "segmentation":
|
| 52 |
+
models[model_name] = pipeline("image-segmentation", model="nvidia/segformer-b0-finetuned-ade-512-512")
|
| 53 |
+
elif model_name == "captioning":
|
| 54 |
+
models[model_name] = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base")
|
| 55 |
+
elif model_name == "vqa":
|
| 56 |
+
models[model_name] = pipeline("visual-question-answering", model="dandelin/vilt-b32-finetuned-vqa")
|
| 57 |
+
elif model_name == "clip":
|
| 58 |
+
models[model_name] = CLIPModel.from_pretrained("openai/clip-vit-base-patch32")
|
| 59 |
+
models[f"{model_name}_processor"] = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
|
| 60 |
+
return models[model_name]
|
| 61 |
+
|
| 62 |
+
# Функции для обработки аудио
|
| 63 |
+
def audio_classification(audio_file, model_type):
|
| 64 |
+
classifier = load_audio_model(model_type)
|
| 65 |
+
results = classifier(audio_file)
|
| 66 |
+
|
| 67 |
+
output = "Топ-5 предсказаний:\n"
|
| 68 |
+
for i, result in enumerate(results[:5]):
|
| 69 |
+
output += f"{i+1}. {result['label']}: {result['score']:.4f}\n"
|
| 70 |
+
|
| 71 |
+
return output
|
| 72 |
+
|
| 73 |
+
def speech_recognition(audio_file, model_type):
|
| 74 |
+
asr_pipeline = load_audio_model(model_type)
|
| 75 |
+
|
| 76 |
+
if model_type == "whisper":
|
| 77 |
+
result = asr_pipeline(audio_file, generate_kwargs={"language": "russian"})
|
| 78 |
+
else:
|
| 79 |
+
result = asr_pipeline(audio_file)
|
| 80 |
+
|
| 81 |
+
return result['text']
|
| 82 |
+
|
| 83 |
+
def text_to_speech(text, model_type):
|
| 84 |
+
if model_type == "silero":
|
| 85 |
+
# Silero TTS
|
| 86 |
+
model, _ = torch.hub.load(repo_or_dir='snakers4/silero-models',
|
| 87 |
+
model='silero_tts',
|
| 88 |
+
language='ru',
|
| 89 |
+
speaker='ru_v3')
|
| 90 |
+
|
| 91 |
+
with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as f:
|
| 92 |
+
model.save_wav(text=text, speaker='aidar', sample_rate=48000, audio_path=f.name)
|
| 93 |
+
return f.name
|
| 94 |
+
|
| 95 |
+
elif model_type == "gtts":
|
| 96 |
+
# Google TTS
|
| 97 |
+
with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as f:
|
| 98 |
+
tts = gTTS(text=text, lang='ru')
|
| 99 |
+
tts.save(f.name)
|
| 100 |
+
return f.name
|
| 101 |
+
|
| 102 |
+
elif model_type == "mms":
|
| 103 |
+
# Facebook MMS TTS
|
| 104 |
+
model = VitsModel.from_pretrained("facebook/mms-tts-rus")
|
| 105 |
+
tokenizer = AutoTokenizer.from_pretrained("facebook/mms-tts-rus")
|
| 106 |
+
|
| 107 |
+
inputs = tokenizer(text, return_tensors="pt")
|
| 108 |
+
with torch.no_grad():
|
| 109 |
+
output = model(**inputs).waveform
|
| 110 |
+
|
| 111 |
+
with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as f:
|
| 112 |
+
sf.write(f.name, output.numpy().squeeze(), model.config.sampling_rate)
|
| 113 |
+
return f.name
|
| 114 |
+
|
| 115 |
+
# Функции для обработки изображений
|
| 116 |
+
def object_detection(image):
|
| 117 |
+
detector = load_image_model("object_detection")
|
| 118 |
+
results = detector(image)
|
| 119 |
+
|
| 120 |
+
# Рисуем bounding boxes
|
| 121 |
+
draw = ImageDraw.Draw(image)
|
| 122 |
+
for result in results:
|
| 123 |
+
box = result['box']
|
| 124 |
+
label = result['label']
|
| 125 |
+
score = result['score']
|
| 126 |
+
|
| 127 |
+
draw.rectangle([box['xmin'], box['ymin'], box['xmax'], box['ymax']],
|
| 128 |
+
outline='red', width=3)
|
| 129 |
+
draw.text((box['xmin'], box['ymin']),
|
| 130 |
+
f"{label}: {score:.2f}", fill='red')
|
| 131 |
+
|
| 132 |
+
return image
|
| 133 |
+
|
| 134 |
+
def image_segmentation(image):
|
| 135 |
+
segmenter = load_image_model("segmentation")
|
| 136 |
+
results = segmenter(image)
|
| 137 |
+
|
| 138 |
+
# Возвращаем первую маску сегментации
|
| 139 |
+
return results[0]['mask']
|
| 140 |
+
|
| 141 |
+
def image_captioning(image):
|
| 142 |
+
captioner = load_image_model("captioning")
|
| 143 |
+
result = captioner(image)
|
| 144 |
+
return result[0]['generated_text']
|
| 145 |
+
|
| 146 |
+
def visual_question_answering(image, question):
|
| 147 |
+
vqa_pipeline = load_image_model("vqa")
|
| 148 |
+
result = vqa_pipeline(image, question)
|
| 149 |
+
return f"{result[0]['answer']} (confidence: {result[0]['score']:.3f})"
|
| 150 |
+
|
| 151 |
+
def zero_shot_classification(image, classes):
|
| 152 |
+
model = load_image_model("clip")
|
| 153 |
+
processor = models["clip_processor"]
|
| 154 |
+
|
| 155 |
+
class_list = [cls.strip() for cls in classes.split(",")]
|
| 156 |
+
|
| 157 |
+
inputs = processor(text=class_list, images=image, return_tensors="pt", padding=True)
|
| 158 |
+
with torch.no_grad():
|
| 159 |
+
outputs = model(**inputs)
|
| 160 |
+
logits_per_image = outputs.logits_per_image
|
| 161 |
+
probs = logits_per_image.softmax(dim=1)
|
| 162 |
+
|
| 163 |
+
result = "Zero-Shot Classification Results:\n"
|
| 164 |
+
for i, cls in enumerate(class_list):
|
| 165 |
+
result += f"{cls}: {probs[0][i].item():.4f}\n"
|
| 166 |
+
|
| 167 |
+
return result
|
| 168 |
+
|
| 169 |
+
def image_retrieval(images, query):
|
| 170 |
+
if not images or not query:
|
| 171 |
+
return "Пожалуйста, загрузите изображения и введите запрос"
|
| 172 |
+
|
| 173 |
+
# Используем CLIP для поиска
|
| 174 |
+
model = load_image_model("clip")
|
| 175 |
+
processor = models["clip_processor"]
|
| 176 |
+
|
| 177 |
+
# Обрабатываем все изображения
|
| 178 |
+
image_inputs = processor(images=images, return_tensors="pt", padding=True)
|
| 179 |
+
with torch.no_grad():
|
| 180 |
+
image_embeddings = model.get_image_features(**image_inputs)
|
| 181 |
+
image_embeddings = image_embeddings / image_embeddings.norm(dim=-1, keepdim=True)
|
| 182 |
+
|
| 183 |
+
# Обрабатываем текстовый запрос
|
| 184 |
+
text_inputs = processor(text=[query], return_tensors="pt", padding=True)
|
| 185 |
+
with torch.no_grad():
|
| 186 |
+
text_embeddings = model.get_text_features(**text_inputs)
|
| 187 |
+
text_embeddings = text_embeddings / text_embeddings.norm(dim=-1, keepdim=True)
|
| 188 |
+
|
| 189 |
+
# Вычисляем схожести
|
| 190 |
+
similarities = (image_embeddings @ text_embeddings.T)
|
| 191 |
+
|
| 192 |
+
# Находим лучшее изображение
|
| 193 |
+
best_idx = similarities.argmax().item()
|
| 194 |
+
best_score = similarities[best_idx].item()
|
| 195 |
+
|
| 196 |
+
return f"Лучшее изображение: #{best_idx + 1} (схожесть: {best_score:.4f})", images[best_idx]
|
| 197 |
+
|
| 198 |
+
# Создаем интерфейс Gradio
|
| 199 |
+
with gr.Blocks(title="Multimodal AI Demo", theme=gr.themes.Soft()) as demo:
|
| 200 |
+
gr.Markdown("# 🎯 Мультимодальные AI модели")
|
| 201 |
+
gr.Markdown("Демонстрация различных задач компьютерного зрения и обработки звука с использованием Hugging Face Transformers")
|
| 202 |
+
|
| 203 |
+
with gr.Tab("🎵 Классификация аудио"):
|
| 204 |
+
gr.Markdown("## Zero-Shot Audio Classification")
|
| 205 |
+
with gr.Row():
|
| 206 |
+
with gr.Column():
|
| 207 |
+
audio_input = gr.Audio(label="Загрузите аудиофайл", type="filepath")
|
| 208 |
+
audio_model_dropdown = gr.Dropdown(
|
| 209 |
+
choices=["audio_classifier", "emotion_classifier"],
|
| 210 |
+
label="Выберите модель",
|
| 211 |
+
value="audio_classifier",
|
| 212 |
+
info="audio_classifier - общая классификация, emotion_classifier - эмоции в речи"
|
| 213 |
+
)
|
| 214 |
+
classify_btn = gr.Button("Классифицировать")
|
| 215 |
+
with gr.Column():
|
| 216 |
+
audio_output = gr.Textbox(label="Результаты классификации", lines=10)
|
| 217 |
+
|
| 218 |
+
classify_btn.click(
|
| 219 |
+
fn=audio_classification,
|
| 220 |
+
inputs=[audio_input, audio_model_dropdown],
|
| 221 |
+
outputs=audio_output
|
| 222 |
+
)
|
| 223 |
+
|
| 224 |
+
with gr.Tab("🗣️ Распознавание речи"):
|
| 225 |
+
gr.Markdown("## Automatic Speech Recognition (ASR)")
|
| 226 |
+
with gr.Row():
|
| 227 |
+
with gr.Column():
|
| 228 |
+
asr_audio_input = gr.Audio(label="Загрузите аудио с речью", type="filepath")
|
| 229 |
+
asr_model_dropdown = gr.Dropdown(
|
| 230 |
+
choices=["whisper", "wav2vec2"],
|
| 231 |
+
label="Выберите модель",
|
| 232 |
+
value="whisper",
|
| 233 |
+
info="whisper - многоязычная, wav2vec2 - специализированная для русского"
|
| 234 |
+
)
|
| 235 |
+
transcribe_btn = gr.Button("Транскрибировать")
|
| 236 |
+
with gr.Column():
|
| 237 |
+
asr_output = gr.Textbox(label="Транскрипция", lines=5)
|
| 238 |
+
|
| 239 |
+
transcribe_btn.click(
|
| 240 |
+
fn=speech_recognition,
|
| 241 |
+
inputs=[asr_audio_input, asr_model_dropdown],
|
| 242 |
+
outputs=asr_output
|
| 243 |
+
)
|
| 244 |
+
|
| 245 |
+
with gr.Tab("🔊 Синтез речи"):
|
| 246 |
+
gr.Markdown("## Text-to-Speech (TTS)")
|
| 247 |
+
with gr.Row():
|
| 248 |
+
with gr.Column():
|
| 249 |
+
tts_text_input = gr.Textbox(
|
| 250 |
+
label="Введите текст для синтеза",
|
| 251 |
+
placeholder="Введите текст на русском языке...",
|
| 252 |
+
lines=3
|
| 253 |
+
)
|
| 254 |
+
tts_model_dropdown = gr.Dropdown(
|
| 255 |
+
choices=["silero", "gtts", "mms"],
|
| 256 |
+
label="Выберите модель",
|
| 257 |
+
value="silero",
|
| 258 |
+
info="silero - высокое качество, gtts - Google TTS, mms - Facebook MMS"
|
| 259 |
+
)
|
| 260 |
+
synthesize_btn = gr.Button("Синтезировать речь")
|
| 261 |
+
with gr.Column():
|
| 262 |
+
tts_output = gr.Audio(label="Синтезированная речь")
|
| 263 |
+
|
| 264 |
+
synthesize_btn.click(
|
| 265 |
+
fn=text_to_speech,
|
| 266 |
+
inputs=[tts_text_input, tts_model_dropdown],
|
| 267 |
+
outputs=tts_output
|
| 268 |
+
)
|
| 269 |
+
|
| 270 |
+
with gr.Tab("📦 Детекция объектов"):
|
| 271 |
+
gr.Markdown("## Object Detection")
|
| 272 |
+
with gr.Row():
|
| 273 |
+
with gr.Column():
|
| 274 |
+
obj_detection_input = gr.Image(label="Загрузите изображение", type="pil")
|
| 275 |
+
detect_btn = gr.Button("Обнаружить объекты")
|
| 276 |
+
with gr.Column():
|
| 277 |
+
obj_detection_output = gr.Image(label="Результат детекции")
|
| 278 |
+
|
| 279 |
+
detect_btn.click(
|
| 280 |
+
fn=object_detection,
|
| 281 |
+
inputs=obj_detection_input,
|
| 282 |
+
outputs=obj_detection_output
|
| 283 |
+
)
|
| 284 |
+
|
| 285 |
+
with gr.Tab("🎨 Сегментация"):
|
| 286 |
+
gr.Markdown("## Image Segmentation")
|
| 287 |
+
with gr.Row():
|
| 288 |
+
with gr.Column():
|
| 289 |
+
seg_input = gr.Image(label="Загрузите изображение", type="pil")
|
| 290 |
+
segment_btn = gr.Button("Сегментировать")
|
| 291 |
+
with gr.Column():
|
| 292 |
+
seg_output = gr.Image(label="Маска сегментации")
|
| 293 |
+
|
| 294 |
+
segment_btn.click(
|
| 295 |
+
fn=image_segmentation,
|
| 296 |
+
inputs=seg_input,
|
| 297 |
+
outputs=seg_output
|
| 298 |
+
)
|
| 299 |
+
|
| 300 |
+
with gr.Tab("📝 Описание изображений"):
|
| 301 |
+
gr.Markdown("## Image Captioning")
|
| 302 |
+
with gr.Row():
|
| 303 |
+
with gr.Column():
|
| 304 |
+
caption_input = gr.Image(label="Загрузите изображение", type="pil")
|
| 305 |
+
caption_btn = gr.Button("Сгенерировать описание")
|
| 306 |
+
with gr.Column():
|
| 307 |
+
caption_output = gr.Textbox(label="Описание изображения", lines=3)
|
| 308 |
+
|
| 309 |
+
caption_btn.click(
|
| 310 |
+
fn=image_captioning,
|
| 311 |
+
inputs=caption_input,
|
| 312 |
+
outputs=caption_output
|
| 313 |
+
)
|
| 314 |
+
|
| 315 |
+
with gr.Tab("❓ Визуальные вопросы"):
|
| 316 |
+
gr.Markdown("## Visual Question Answering")
|
| 317 |
+
with gr.Row():
|
| 318 |
+
with gr.Column():
|
| 319 |
+
vqa_image_input = gr.Image(label="Загрузите изображение", type="pil")
|
| 320 |
+
vqa_question_input = gr.Textbox(
|
| 321 |
+
label="Вопрос об изображении",
|
| 322 |
+
placeholder="Что происходит на этом изображении?",
|
| 323 |
+
lines=2
|
| 324 |
+
)
|
| 325 |
+
vqa_btn = gr.Button("Ответить на вопрос")
|
| 326 |
+
with gr.Column():
|
| 327 |
+
vqa_output = gr.Textbox(label="Ответ", lines=3)
|
| 328 |
+
|
| 329 |
+
vqa_btn.click(
|
| 330 |
+
fn=visual_question_answering,
|
| 331 |
+
inputs=[vqa_image_input, vqa_question_input],
|
| 332 |
+
outputs=vqa_output
|
| 333 |
+
)
|
| 334 |
+
|
| 335 |
+
with gr.Tab("🎯 Zero-Shot классификация"):
|
| 336 |
+
gr.Markdown("## Zero-Shot Image Classification")
|
| 337 |
+
with gr.Row():
|
| 338 |
+
with gr.Column():
|
| 339 |
+
zs_image_input = gr.Image(label="Загрузите изображение", type="pil")
|
| 340 |
+
zs_classes_input = gr.Textbox(
|
| 341 |
+
label="Классы для классификации (через запятую)",
|
| 342 |
+
placeholder="человек, машина, дерево, здание, животное",
|
| 343 |
+
lines=2
|
| 344 |
+
)
|
| 345 |
+
zs_classify_btn = gr.Button("Классифицировать")
|
| 346 |
+
with gr.Column():
|
| 347 |
+
zs_output = gr.Textbox(label="Результаты классификации", lines=10)
|
| 348 |
+
|
| 349 |
+
zs_classify_btn.click(
|
| 350 |
+
fn=zero_shot_classification,
|
| 351 |
+
inputs=[zs_image_input, zs_classes_input],
|
| 352 |
+
outputs=zs_output
|
| 353 |
+
)
|
| 354 |
+
|
| 355 |
+
with gr.Tab("🔍 Поиск изображений"):
|
| 356 |
+
gr.Markdown("## Image Retrieval")
|
| 357 |
+
with gr.Row():
|
| 358 |
+
with gr.Column():
|
| 359 |
+
retrieval_images_input = gr.Gallery(
|
| 360 |
+
label="Загрузите изображения для поиска",
|
| 361 |
+
type="pil"
|
| 362 |
+
)
|
| 363 |
+
retrieval_query_input = gr.Textbox(
|
| 364 |
+
label="Текстовый запрос",
|
| 365 |
+
placeholder="описание того, что вы ищете...",
|
| 366 |
+
lines=2
|
| 367 |
+
)
|
| 368 |
+
retrieval_btn = gr.Button("Найти изображение")
|
| 369 |
+
with gr.Column():
|
| 370 |
+
retrieval_output_text = gr.Textbox(label="Результат поиска")
|
| 371 |
+
retrieval_output_image = gr.Image(label="Найденное изображение")
|
| 372 |
+
|
| 373 |
+
retrieval_btn.click(
|
| 374 |
+
fn=image_retrieval,
|
| 375 |
+
inputs=[retrieval_images_input, retrieval_query_input],
|
| 376 |
+
outputs=[retrieval_output_text, retrieval_output_image]
|
| 377 |
+
)
|
| 378 |
+
|
| 379 |
+
gr.Markdown("---")
|
| 380 |
+
gr.Markdown("### 📊 Поддерживаемые задачи:")
|
| 381 |
+
gr.Markdown("""
|
| 382 |
+
- **🎵 Аудио**: Классификация, распознавание речи, синтез речи
|
| 383 |
+
- **👁️ Компьютерное зрение**: Детекция объектов, сегментация, описание изображений
|
| 384 |
+
- **🤖 Мультимодальные**: Визуальные вопросы, zero-shot классификация, поиск по изображениям
|
| 385 |
+
""")
|
| 386 |
+
|
| 387 |
+
if __name__ == "__main__":
|
| 388 |
+
demo.launch(share=True)
|
requirements.txt
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
torch>=2.0.0
|
| 2 |
+
torchaudio>=2.0.0
|
| 3 |
+
transformers>=4.30.0
|
| 4 |
+
gradio>=4.0.0
|
| 5 |
+
pillow>=9.0.0
|
| 6 |
+
numpy>=1.21.0
|
| 7 |
+
soundfile>=0.12.0
|
| 8 |
+
gtts>=2.3.0
|
| 9 |
+
sentence-transformers>=2.2.0
|
| 10 |
+
librosa>=0.10.0
|
| 11 |
+
requests>=2.28.0
|
| 12 |
+
accelerate>=0.20.0
|