Spaces:
Sleeping
Sleeping
File size: 9,708 Bytes
6d6b8af |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 |
"""
Codette Multimodal Analyzer
Provides lightweight, dependency-safe analysis helpers for text, image,
audio and video inputs. Functions accept either raw bytes, a filesystem
path (str) or simple Python objects (e.g. numpy arrays for audio).
This module intentionally avoids heavy mandatory dependencies. If
optional libraries (Pillow, OpenCV) are installed, the analyzer will use
them for richer metadata extraction; otherwise it falls back to safe
heuristics and headers.
"""
from typing import Dict, Any, List, Union, Optional
import io
import os
import imghdr
import mimetypes
import wave
import struct
try:
from PIL import Image
except Exception:
Image = None
try:
import cv2
except Exception:
cv2 = None
try:
import numpy as np
except Exception:
np = None
class MultimodalAnalyzer:
def __init__(self):
self.supported_modalities = {
"text": self._analyze_text,
"image": self._analyze_image,
"audio": self._analyze_audio,
"video": self._analyze_video,
}
def analyze_content(self, content: Dict[str, Any]) -> Dict[str, Any]:
results: Dict[str, Any] = {}
for modality, data in content.items():
handler = self.supported_modalities.get(modality)
if handler is None:
results[modality] = {"error": "Unsupported modality"}
continue
try:
results[modality] = handler(data)
except Exception as e:
results[modality] = {"error": str(e)}
return results
def _analyze_text(self, text: Union[str, bytes]) -> Dict[str, Any]:
if isinstance(text, bytes):
try:
text = text.decode("utf-8", errors="replace")
except Exception:
text = str(text)
text = text or ""
words = [w for w in text.split() if w]
unique_words = set(w.strip(".,!?;:\"()[]{}") for w in words)
avg_word_len = sum(len(w) for w in words) / len(words) if words else 0
has_questions = "?" in text
has_exclamations = "!" in text
language = "en" if all(ord(c) < 128 for c in text) else "non-en"
return {
"type": "text",
"length": len(text),
"word_count": len(words),
"unique_word_count": len(unique_words),
"avg_word_length": round(avg_word_len, 2),
"has_content": bool(text.strip()),
"has_questions": has_questions,
"has_exclamations": has_exclamations,
"language_estimate": language,
}
def _read_bytes_or_path(self, data: Union[bytes, str]) -> Optional[bytes]:
if data is None:
return None
if isinstance(data, bytes):
return data
if isinstance(data, str):
if os.path.exists(data):
try:
with open(data, "rb") as f:
return f.read()
except Exception:
return None
# treat string as raw small payload
return data.encode("utf-8", errors="replace")
return None
def _analyze_image(self, image_data: Union[bytes, str, None]) -> Dict[str, Any]:
raw = self._read_bytes_or_path(image_data)
info: Dict[str, Any] = {"type": "image", "has_content": bool(raw)}
if not raw:
return info
fmt = None
try:
fmt = imghdr.what(None, h=raw)
except Exception:
fmt = None
if fmt:
info["format"] = fmt
else:
# fallback to mime type by filename if provided
if isinstance(image_data, str):
mt, _ = mimetypes.guess_type(image_data)
info["format"] = mt or "unknown"
else:
info["format"] = "unknown"
if Image is not None:
try:
img = Image.open(io.BytesIO(raw))
info.update({
"width": img.width,
"height": img.height,
"mode": img.mode,
"has_alpha": "A" in img.getbands(),
})
img.close()
except Exception:
pass
return info
def _analyze_audio(self, audio_data: Union[bytes, str, Any, None]) -> Dict[str, Any]:
info: Dict[str, Any] = {"type": "audio", "has_content": False, "format": "unknown"}
if audio_data is None:
return info
if np is not None and isinstance(audio_data, np.ndarray):
arr = audio_data
info["has_content"] = getattr(arr, "size", 0) > 0
info["format"] = "numpy.ndarray"
try:
samples = arr.astype(float)
rms = float(np.sqrt(np.mean(samples ** 2)))
info["rms"] = float(rms)
info["duration_seconds_estimate"] = None
except Exception:
pass
return info
raw = self._read_bytes_or_path(audio_data)
if not raw:
return info
info["has_content"] = True
# Try WAV detection
try:
bio = io.BytesIO(raw)
with wave.open(bio, "rb") as w:
nchannels = w.getnchannels()
sampwidth = w.getsampwidth()
framerate = w.getframerate()
nframes = w.getnframes()
duration = nframes / float(framerate) if framerate else None
info.update({
"format": "wav",
"channels": nchannels,
"sample_width": sampwidth,
"frame_rate": framerate,
"n_frames": nframes,
"duration_seconds": duration,
})
try:
frames = w.readframes(min(nframes, 44100))
if frames:
# unpack frames to numpy array for RMS
if sampwidth == 1:
dtype = np.uint8
elif sampwidth == 2:
dtype = np.int16
elif sampwidth == 4:
dtype = np.int32
else:
dtype = np.int16
samples = np.frombuffer(frames, dtype=dtype).astype(float)
if nchannels > 1:
samples = samples.reshape(-1, nchannels)
samples = samples.mean(axis=1)
rms = float(np.sqrt(np.mean((samples) ** 2)))
info["rms"] = rms
except Exception:
pass
return info
except wave.Error:
pass
except Exception:
pass
# fallback: try to guess mime type by filename
if isinstance(audio_data, str):
mt, _ = mimetypes.guess_type(audio_data)
if mt:
info["format"] = mt
return info
def _analyze_video(self, video_data: Union[bytes, str, None]) -> Dict[str, Any]:
info: Dict[str, Any] = {"type": "video", "has_content": False, "format": "unknown"}
raw = self._read_bytes_or_path(video_data)
if not raw:
return info
info["has_content"] = True
if isinstance(video_data, str):
mt, _ = mimetypes.guess_type(video_data)
if mt:
info["format"] = mt
# If OpenCV is available, try to extract metadata
if cv2 is not None and isinstance(video_data, str) and os.path.exists(video_data):
try:
cap = cv2.VideoCapture(video_data)
if cap.isOpened():
fps = cap.get(cv2.CAP_PROP_FPS) or None
frame_count = int(cap.get(cv2.CAP_PROP_FRAME_COUNT) or 0)
width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH) or 0)
height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT) or 0)
duration = frame_count / fps if fps else None
info.update({
"fps": fps,
"frame_count": frame_count,
"width": width,
"height": height,
"duration_seconds": duration,
})
cap.release()
except Exception:
pass
return info
def combine_modalities(self, analyses: Dict[str, Any]) -> Dict[str, Any]:
modalities_present = [k for k, v in analyses.items() if not v.get("error")]
summary = {
"modalities_present": modalities_present,
"modality_count": len(modalities_present),
"complete_analysis": all(not v.get("error") for v in analyses.values()),
"analyses": analyses,
}
if "text" in analyses and "image" in analyses:
t = analyses.get("text", {})
img = analyses.get("image", {})
summary["text_and_image"] = {
"text_length": t.get("length"),
"image_size": (img.get("width"), img.get("height")),
}
return summary
def get_supported_modalities(self) -> List[str]:
return list(self.supported_modalities.keys()) |