File size: 9,054 Bytes
ba1d61a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
import os
import json
import argparse
import base64
import cv2
import tempfile
import subprocess
import shutil
import io
import time
import glob
from openai import OpenAI
from tqdm import tqdm
from PIL import Image

GENERIC_RESULT_PATTERN = "_result.json"

def get_media_type(file_path: str) -> str:
    ext = os.path.splitext(file_path)[1].lower()
    if ext in ['.mp4', '.avi', '.mov', '.mkv', '.webm']:
        return 'video'
    elif ext in ['.jpg', '.jpeg', '.png', '.bmp', '.gif', '.webp']:
        return 'image'
    else:
        raise ValueError(f"Unsupported file format: {ext}")

def encode_media_to_base64(media_path: str) -> str:
    try:
        with open(media_path, "rb") as media_file:
            return base64.b64encode(media_file.read()).decode('utf-8')
    except FileNotFoundError:
        raise
    except Exception as e:
        raise IOError(f"Could not read or encode file {media_path}: {e}")

def extract_keyframes(video_path: str, max_frames: int = 20) -> list:
    cap = cv2.VideoCapture(video_path)
    if not cap.isOpened():
        raise ValueError(f"Cannot open video file: {video_path}")

    total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
    video_fps = cap.get(cv2.CAP_PROP_FPS)
    duration = total_frames / video_fps if video_fps > 0 else 0

    if duration <= 20:
        target_frames = min(max_frames, int(duration))
        frame_interval = max(1, int(video_fps))  # 1 frame per second
    else:
        target_frames = max_frames
        frame_interval = max(1, int(total_frames / max_frames))


    keyframes = []
    frame_count = 0
    sampled_count = 0

    while cap.isOpened() and sampled_count < target_frames:
        ret, frame = cap.read()
        if not ret:
            break

        if frame_count % frame_interval == 0:
            frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
            pil_image = Image.fromarray(frame_rgb)
            keyframes.append(pil_image)
            sampled_count += 1

        frame_count += 1

    cap.release()
    return keyframes

def extract_audio_to_text(video_path: str, client: OpenAI) -> str:
    try:
        with tempfile.NamedTemporaryFile(suffix=".mp3", delete=False) as temp_audio:
            temp_audio_path = temp_audio.name
        command = [
            'ffmpeg', '-i', video_path, '-vn', '-acodec', 'mp3',
            '-ar', '16000', '-ac', '1', '-y', temp_audio_path
        ]
        result = subprocess.run(command, capture_output=True, text=True)

        if result.returncode != 0:
            print(f"FFmpeg audio extraction failed: {result.stderr}")
            os.unlink(temp_audio_path)
            return "Audio extraction failed."
        with open(temp_audio_path, 'rb') as audio_file:

            transcription = client.audio.transcriptions.create(
                model="whisper-1",
                file=(os.path.basename(temp_audio_path), audio_file.read()),
                response_format="text"
            )
        
        os.unlink(temp_audio_path)
        return transcription

    except Exception as e:
        print(f"Audio processing failed: {e}")
 
        if 'temp_audio_path' in locals() and os.path.exists(temp_audio_path):
            os.unlink(temp_audio_path)
        return "Audio processing failed."

def process_single_sample(client: OpenAI, media_full_path: str, prompt_text: str) -> str:

    clean_prompt = prompt_text.replace("<image>", "").replace("<video>", "").strip()
    media_type = get_media_type(media_full_path)

    try:
        content = [{"type": "text", "text": clean_prompt}]

        if media_type == 'image':
            base64_media = encode_media_to_base64(media_full_path)
            content.append({"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{base64_media}"}})
        
        elif media_type == 'video':
            key_frames = extract_keyframes(media_full_path)
            audio_text = ""
            if shutil.which('ffmpeg'):
                try:
                    audio_text = extract_audio_to_text(media_full_path, client)
                except Exception as audio_e:
                    print(f" Could not extract audio from video: {audio_e}")
            else:
                print(" ffmpeg not found, skipping audio extraction.")

            for frame in key_frames:
                with io.BytesIO() as buffer:
                    frame.save(buffer, format='JPEG', quality=85)
                    frame_base64 = base64.b64encode(buffer.getvalue()).decode('utf-8')
                    content.append({"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{frame_base64}"}})
            
            if audio_text:
                full_prompt_with_audio = f"{clean_prompt}\n\n--- Video Audio Transcription ---\n{audio_text}"
                content[0] = {"type": "text", "text": full_prompt_with_audio}

        messages = [{"role": "user", "content": content}]
        response = client.chat.completions.create(
            model="gpt-4o",
            messages=messages,
            max_tokens=1024,
            temperature=0,
            timeout=120.0
        )
        return response.choices[0].message.content

    except Exception as e:
        try:
            messages = [{"role": "user", "content": clean_prompt}]
            response = client.chat.completions.create(
                model="gpt-4o",
                messages=messages,
                max_tokens=1024,
                temperature=0,
                timeout=60.0
            )
            return response.choices[0].message.content
        except Exception as fallback_e:
            print(f"  Text-only fallback also failed: {fallback_e}")
            return f"ERROR: Fallback failed. Details: {fallback_e}"

def process_dataset_file(dataset_path: str, client: OpenAI, result_suffix: str):

    result_path = f"{os.path.splitext(dataset_path)[0]}{result_suffix}"
    if os.path.exists(result_path):
        print(f"Result file '{os.path.basename(result_path)}' already exists, skipping.")
        return

    try:
        with open(dataset_path, 'r', encoding='utf-8') as f:
            data = json.load(f)
    except (json.JSONDecodeError, FileNotFoundError) as e:
        print(f" Could not read or parse JSON file {dataset_path}: {e}")
        return

    all_results = []
    base_dir = os.path.dirname(dataset_path)

    for item in tqdm(data, desc=f"  Processing {os.path.basename(dataset_path)}"):
        start_time = time.time()
        model_output = ""
        try:
            prompt = item['conversations'][0]['value']
            ground_truth = item['conversations'][1]['value']
            media_path_key = 'image' if 'image' in item else 'video'
            media_relative_path = item.get(media_path_key)

            if not media_relative_path:
                raise ValueError("JSON item is missing 'image' or 'video' key.")
            
            media_full_path = os.path.join(base_dir, media_relative_path)
            if not os.path.exists(media_full_path):
                raise FileNotFoundError(f"Media file not found: {media_full_path}")
            
            model_output = process_single_sample(client, media_full_path, prompt)
            
        except Exception as e:
            model_output = f"ERROR: {str(e)}"
            print(f" Failed to process item {item.get('id', 'N/A')}: {e}")
        
        end_time = time.time()
        all_results.append({
            "id": item.get('id', 'N/A'), 
            "prompt": prompt, 
            "model_output": model_output, 
            "ground_truth": ground_truth, 
            "processing_time_seconds": round(end_time - start_time, 2)
        })

    with open(result_path, 'w', encoding='utf-8') as f:
        json.dump(all_results, f, indent=4, ensure_ascii=False)
    print(f" Processing complete. Results saved to: {result_path}")

def main():
    parser = argparse.ArgumentParser(description="Run inference on a dataset using GPT-4o with multimodal capabilities.")
    parser.add_argument("--api-key", required=True, help="OpenAI API key.")
    parser.add_argument("--model-name", default="gpt-4o", help="The model to use (default: gpt-4o).")
    parser.add_argument("--result-suffix", default="_gpt4o_result.json", help="Suffix for the output result files.")
    
    args = parser.parse_args()

    try:
        client = OpenAI(api_key=args.api_key)
    except Exception as e:
        print(f"[Could not initialize OpenAI client: {e}")
        return

    if not shutil.which('ffmpeg'):
        print("ffmpeg not found. Video audio extraction will be disabled.")

    source_json_files = [
        f for f in glob.glob('*.json') 
        if not f.endswith(args.result_suffix) and GENERIC_RESULT_PATTERN not in f
    ]

    if not source_json_files:
        print("\n No source JSON files.")
    else:
        print(f"\nFound {len(source_json_files)} dataset(s) to process.")

    for json_file in sorted(source_json_files):
        process_dataset_file(json_file, client, args.result_suffix)
            
if __name__ == "__main__":
    main()