Spaces:
Sleeping
Sleeping
File size: 17,611 Bytes
3d0817d 60e56c1 3d0817d | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 | import streamlit as st
import numpy as np
import tensorflow as tf
from tensorflow.keras.applications.vgg16 import preprocess_input as vgg_preprocess
from tensorflow.keras.applications.resnet50 import preprocess_input as resnet_preprocess
from tensorflow.keras.applications.efficientnet import preprocess_input as eff_preprocess
from PIL import Image
import os
import time
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
# CLASS NAMES
# Hardcoded from training notebooks (flow_from_directory sorts alphabetically)
# This is the exact order your models learned โ no class_indices.json needed.
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
CLASS_NAMES = [
'airplane', 'bench', 'bicycle', 'bird', 'bottle',
'bowl', 'bus', 'cake', 'car', 'cat',
'chair', 'couch', 'cow', 'cup', 'dog',
'elephant', 'horse', 'motorcycle', 'person', 'pizza',
'potted plant', 'stop sign', 'traffic light', 'train', 'truck',
]
# Note: flow_from_directory loads classes in alphabetical order.
# The list above is sorted AโZ to match exactly what your models output.
CLASS_ICONS = {
'airplane': 'โ๏ธ', 'bench': '๐ช', 'bicycle': '๐ฒ', 'bird': '๐ฆ',
'bottle': '๐ถ', 'bowl': '๐ฅฃ', 'bus': '๐', 'cake': '๐',
'car': '๐', 'cat': '๐ฑ', 'chair': '๐ช', 'couch': '๐๏ธ',
'cow': '๐ฎ', 'cup': 'โ', 'dog': '๐ถ', 'elephant': '๐',
'horse': '๐ด', 'motorcycle': '๐๏ธ', 'person': '๐ง', 'pizza': '๐',
'potted plant': '๐ชด', 'stop sign': '๐', 'traffic light': '๐ฆ',
'train': '๐', 'truck': '๐',
}
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
# MODEL CONFIGS
# Preprocessing matches exactly what each training notebook used
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
# PLOT_DIR = "models/result"
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
PLOT_DIR = os.path.join(BASE_DIR, "models", "result")
MODEL_CONFIGS = {
"EfficientNetB0": {
"path": os.path.join(PLOT_DIR, "efficientnetb0_best.keras"),
"preprocess": eff_preprocess, # Phase 2.4: preprocess_input from efficientnet
"rescale": False,
"color": "#f39c12",
"icon": "๐จ",
"accuracy": "92%",
"speed": "80 ms",
"description": "Highest accuracy ยท Compound scaling",
},
"ResNet50": {
"path": os.path.join(PLOT_DIR, "resnet50_best.keras"),
"preprocess": resnet_preprocess, # Phase 2.2: preprocess_input from resnet50
"rescale": False,
"color": "#e74c3c",
"icon": "๐ฅ",
"accuracy": "88%",
"speed": "100 ms",
"description": "Strong all-rounder ยท Residual learning",
},
"MobileNetV2": {
"path": os.path.join(PLOT_DIR, "mobilenetv2_final.h5"),
"preprocess": None, # Phase 2.3: ImageDataGenerator(rescale=1./255)
"rescale": True,
"color": "#2ecc71",
"icon": "๐ฉ",
"accuracy": "85%",
"speed": "50 ms",
"description": "Fastest ยท Lightweight ยท Edge-ready",
},
"VGG16": {
"path": os.path.join(PLOT_DIR, "vgg16_best.keras"),
"preprocess": vgg_preprocess, # Phase 2.1: preprocess_input from vgg16
"rescale": False,
"color": "#3498db",
"icon": "๐ฆ",
"accuracy": "83%",
"speed": "150 ms",
"description": "Classic CNN ยท Reliable baseline",
},
}
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
# HELPERS
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
@st.cache_resource(show_spinner=False)
def load_all_models():
loaded = {}
for name, cfg in MODEL_CONFIGS.items():
if os.path.exists(cfg["path"]):
try:
loaded[name] = tf.keras.models.load_model(cfg["path"])
except Exception as e:
st.warning(f"โ ๏ธ Failed to load {name}: {e}")
loaded[name] = None
else:
loaded[name] = None
return loaded
def prepare_image(pil_img, cfg):
"""Resize, apply the correct preprocessing, return (1, 224, 224, 3) array."""
img = pil_img.resize((224, 224))
arr = np.array(img, dtype=np.float32)
if cfg["rescale"]:
arr = arr / 255.0 # MobileNetV2 path
else:
arr = cfg["preprocess"](arr) # VGG16 / ResNet50 / EfficientNet path
return np.expand_dims(arr, axis=0)
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
# PAGE HEADER
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
st.title("๐ผ๏ธ Image Classification")
st.caption("Upload any image โ all 4 CNN models classify it simultaneously")
# โโ Model status cards โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
st.markdown("### ๐ค Available Models")
cols = st.columns(4)
for col, (mname, cfg) in zip(cols, MODEL_CONFIGS.items()):
exists = os.path.exists(cfg["path"])
status_label = "โ
Ready" if exists else "โ File missing"
status_color = "#2ecc71" if exists else "#e74c3c"
with col:
st.markdown(
f"""
<div style="
border: 2px solid {cfg['color']}55;
border-radius: 12px; padding: 12px 10px;
background: {cfg['color']}0d; text-align: center; height: 148px;
">
<p style="font-size:22px; margin:0;">{cfg['icon']}</p>
<p style="font-size:13px; font-weight:700; color:{cfg['color']}; margin:4px 0 2px;">{mname}</p>
<p style="font-size:11px; color:#666; margin:0 0 5px;">{cfg['description']}</p>
<p style="font-size:11px; color:#444; margin:0;">๐ฏ {cfg['accuracy']} โก {cfg['speed']}</p>
<p style="font-size:10px; color:{status_color}; margin:5px 0 0; font-weight:600;">{status_label}</p>
</div>
""",
unsafe_allow_html=True,
)
st.divider()
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
# UPLOAD
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
uploaded = st.file_uploader(
"๐ค Upload an image to classify",
type=["jpg", "jpeg", "png"],
help="Best results with a single clear object centred in the frame.",
)
if not uploaded:
st.info("๐ Upload an image above to see predictions from all 4 models.")
st.stop()
image = Image.open(uploaded).convert("RGB")
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
# LOAD MODELS
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
with st.spinner("โณ Loading models (cached after first run)โฆ"):
models = load_all_models()
available = {k: v for k, v in models.items() if v is not None}
if not available:
st.error(
"โ No model files found in `model/result/`. "
"Make sure `.keras` / `.h5` files are present."
)
st.stop()
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
# RUN INFERENCE
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
all_results = {}
bar = st.progress(0, text="Running inferenceโฆ")
for i, (mname, model) in enumerate(available.items()):
cfg = MODEL_CONFIGS[mname]
arr = prepare_image(image, cfg)
t0 = time.time()
preds = model.predict(arr, verbose=0)[0]
ms = (time.time() - t0) * 1000
top5_idx = np.argsort(preds)[::-1][:5]
top5 = [(CLASS_NAMES[j], float(preds[j])) for j in top5_idx]
all_results[mname] = {"top5": top5, "top1": top5[0], "ms": ms, "cfg": cfg}
bar.progress((i + 1) / len(available), text=f"โ
{mname} complete")
bar.empty()
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
# LAYOUT: image | consensus
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
img_col, sum_col = st.columns([1, 1])
with img_col:
st.image(image, caption="๐ท Input Image", use_container_width=True)
st.caption(f"Size: {image.size[0]}ร{image.size[1]} px ยท RGB")
with sum_col:
st.markdown("### ๐ง Model Consensus")
votes = [r["top1"][0] for r in all_results.values()]
consensus = max(set(votes), key=votes.count)
n_agree = votes.count(consensus)
icon = CLASS_ICONS.get(consensus, "๐")
c_color = "#2ecc71" if n_agree >= 3 else "#f39c12" if n_agree == 2 else "#e74c3c"
st.markdown(
f"""
<div style="
background: linear-gradient(135deg, {c_color}22, {c_color}11);
border: 2px solid {c_color}66; border-radius: 14px;
padding: 22px; text-align: center; margin-bottom: 16px;
">
<p style="font-size:52px; margin:0;">{icon}</p>
<p style="font-size:22px; font-weight:800; color:{c_color};
margin:8px 0 4px; text-transform:capitalize;">{consensus}</p>
<p style="font-size:13px; color:#555; margin:0;">{n_agree} / {len(all_results)} models agree</p>
</div>
""",
unsafe_allow_html=True,
)
st.markdown("**Top prediction per model:**")
for mname, res in all_results.items():
cfg = res["cfg"]
label = res["top1"][0]
conf = res["top1"][1]
match = "โ
" if label == consensus else "๐ถ"
st.markdown(
f"""
<div style="
display:flex; justify-content:space-between; align-items:center;
border-left: 4px solid {cfg['color']}; padding: 6px 10px;
margin-bottom: 6px; background: {cfg['color']}0d;
border-radius: 0 8px 8px 0;
">
<span style="font-size:13px; font-weight:600; color:{cfg['color']};">{cfg['icon']} {mname}</span>
<span style="font-size:13px;">{match} {label}</span>
<span style="font-size:12px; color:#888;">{conf:.0%} ยท {res['ms']:.0f}ms</span>
</div>
""",
unsafe_allow_html=True,
)
st.divider()
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
# DETAILED TOP-5 CARDS โ 2-column grid
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
st.markdown("### ๐ Detailed Predictions โ Top 5 Per Model")
names = list(all_results.keys())
for i in range(0, len(names), 2):
row = st.columns(2)
for j, col in enumerate(row):
if i + j >= len(names):
break
mname = names[i + j]
res = all_results[mname]
cfg = res["cfg"]
with col:
# Header
st.markdown(
f"""
<div style="border:2px solid {cfg['color']}66; border-radius:12px 12px 0 0;
padding:10px 14px; background:{cfg['color']}15;">
<span style="font-size:15px; font-weight:700; color:{cfg['color']};">{cfg['icon']} {mname}</span>
<span style="float:right; font-size:12px; color:#666;">โก {res['ms']:.0f} ms</span><br>
<span style="font-size:11px; color:#666;">{cfg['description']}</span>
</div>
""",
unsafe_allow_html=True,
)
# Top-5 bars
for rank, (label, conf) in enumerate(res["top5"]):
li = CLASS_ICONS.get(label, "โข")
bar_c = cfg["color"] if rank == 0 else "#bbb"
bg = f"{cfg['color']}18" if rank == 0 else "transparent"
fw = "700" if rank == 0 else "400"
medal = "๐ฅ" if rank == 0 else f"#{rank+1}"
st.markdown(
f"""
<div style="padding:5px 14px; background:{bg};">
<div style="display:flex; justify-content:space-between; margin-bottom:2px;">
<span style="font-size:13px; font-weight:{fw};">{medal} {li} {label}</span>
<span style="font-size:12px; font-weight:{fw}; color:{bar_c};">{conf:.1%}</span>
</div>
<div style="background:#eee; border-radius:6px; height:7px; overflow:hidden;">
<div style="width:{max(int(conf*100),2)}%; background:{bar_c};
height:100%; border-radius:6px;"></div>
</div>
</div>
""",
unsafe_allow_html=True,
)
# Footer border
st.markdown(
f'<div style="border:2px solid {cfg["color"]}44; border-top:none; '
f'border-radius:0 0 12px 12px; height:8px;"></div>',
unsafe_allow_html=True,
)
st.markdown("")
st.divider()
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
# LIVE INFERENCE TIME โ vertical bars
# โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
st.markdown("### โก Live Inference Time")
max_ms = max(r["ms"] for r in all_results.values()) or 1
speed_cols = st.columns(len(all_results))
for col, (mname, res) in zip(speed_cols, all_results.items()):
cfg = res["cfg"]
pct = int((res["ms"] / max_ms) * 100)
with col:
st.markdown(
f"""
<div style="text-align:center; padding:10px;">
<p style="font-size:12px; font-weight:600; color:{cfg['color']}; margin:0 0 6px;">{cfg['icon']} {mname}</p>
<div style="background:#eee; border-radius:8px; height:80px; position:relative; overflow:hidden;">
<div style="position:absolute; bottom:0; width:100%; height:{pct}%;
background:{cfg['color']}; border-radius:8px;"></div>
</div>
<p style="font-size:13px; font-weight:700; margin:6px 0 0; color:{cfg['color']};">{res['ms']:.0f} ms</p>
</div>
""",
unsafe_allow_html=True,
)
st.caption("๐ Inference times measured live on this machine. GPU will be significantly faster.") |