File size: 10,359 Bytes
c914720
39e0cb0
60fb79c
 
 
39e0cb0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
60fb79c
39e0cb0
 
 
 
 
c914720
39e0cb0
 
 
c914720
39e0cb0
 
 
 
 
 
 
c914720
39e0cb0
 
c914720
 
 
 
 
 
 
 
39e0cb0
 
 
 
60fb79c
39e0cb0
 
 
 
 
60fb79c
39e0cb0
 
 
60fb79c
39e0cb0
 
 
 
60fb79c
c914720
39e0cb0
 
 
60fb79c
39e0cb0
 
 
 
 
 
 
 
 
 
 
 
c914720
39e0cb0
 
 
c914720
39e0cb0
 
c914720
 
 
 
 
 
 
 
 
60fb79c
c914720
39e0cb0
c914720
 
60fb79c
c914720
39e0cb0
 
 
c914720
60fb79c
39e0cb0
60fb79c
39e0cb0
 
 
60fb79c
39e0cb0
 
c914720
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
39e0cb0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c914720
39e0cb0
 
c914720
 
 
 
 
 
 
 
 
39e0cb0
c914720
 
 
 
39e0cb0
c914720
 
39e0cb0
c914720
 
 
39e0cb0
 
 
 
 
 
 
 
 
c914720
39e0cb0
 
 
 
60fb79c
39e0cb0
c914720
60fb79c
39e0cb0
 
 
 
 
 
 
 
 
c914720
 
39e0cb0
 
 
 
 
 
c914720
39e0cb0
 
 
 
 
 
 
 
 
 
 
60fb79c
 
 
 
 
 
c914720
 
 
 
 
 
60fb79c
 
 
 
 
39e0cb0
 
 
 
 
 
 
c914720
 
 
 
 
39e0cb0
 
c914720
39e0cb0
 
 
 
 
 
 
c914720
39e0cb0
 
 
 
 
 
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
import os
from PIL import Image, ImageStat

Image.MAX_IMAGE_PIXELS = 200_000_000

from engine.hoodie_rules import AOP_HOODIE_MAPPING_RULES
from engine.hf_processors import run_optional_ai_processors

def norm(name):
    return name.lower().replace(" ", "_")

def pct_box(w, h, x1, y1, x2, y2):
    return (int(w * x1), int(h * y1), int(w * x2), int(h * y2))

def region_score(img):
    stat = ImageStat.Stat(img.convert("RGB"))
    contrast = sum(stat.stddev) / 3
    saturation_hint = max(stat.mean) - min(stat.mean)
    brightness = sum(stat.mean) / 3
    return contrast + saturation_hint + (255 - abs(140 - brightness)) * 0.12

def analyze_regions(image, prompt, ai_results=None):
    ai_results = ai_results or {}
    w, h = image.size
    p = (prompt or "").lower()

    base = [
        ("PRIMARY_FOCAL_REGION", pct_box(w, h, 0.18, 0.10, 0.82, 0.82)),
        ("SECONDARY_FOCAL_REGION", pct_box(w, h, 0.04, 0.00, 0.75, 0.50)),
        ("ENVIRONMENT_REGION", pct_box(w, h, 0.00, 0.00, 1.00, 1.00)),
        ("LOWER_ANCHOR_REGION", pct_box(w, h, 0.05, 0.58, 0.95, 1.00)),
        ("LEFT_EDGE_FLOW_REGION", pct_box(w, h, 0.00, 0.04, 0.38, 0.96)),
        ("RIGHT_EDGE_FLOW_REGION", pct_box(w, h, 0.62, 0.04, 1.00, 0.96)),
    ]

    regions = []

    for region_type, box in base:
        crop = image.crop(box)
        score = region_score(crop)

        if region_type == "PRIMARY_FOCAL_REGION":
            if any(k in p for k in ["face", "head", "person", "character", "dj", "main", "center", "figure", "subject"]):
                score += 45

        if region_type == "SECONDARY_FOCAL_REGION":
            if any(k in p for k in ["eye", "monster", "upper", "teeth", "dramatic", "background"]):
                score += 35

        if region_type == "LOWER_ANCHOR_REGION":
            if any(k in p for k in ["pocket", "deck", "table", "lower", "base", "ground", "smoke"]):
                score += 30

        if region_type == "ENVIRONMENT_REGION":
            if any(k in p for k in ["pattern", "repeat", "tile", "smoke", "flame", "abstract", "texture"]):
                score += 30

        if region_type == "LEFT_EDGE_FLOW_REGION":
            if any(k in p for k in ["left sleeve", "left edge", "flame", "smoke"]):
                score += 20

        if region_type == "RIGHT_EDGE_FLOW_REGION":
            if any(k in p for k in ["right sleeve", "right edge", "purple", "smoke"]):
                score += 20

        regions.append({
            "type": region_type,
            "box": box,
            "score": round(score, 2),
            "source": "local_region_analysis",
        })

    for category, result in ai_results.items():
        if not result.get("ok"):
            continue

        for det in result.get("detections", []):
            det_type = str(det.get("type", category)).lower()
            box = det.get("box")

            if not box or len(box) != 4:
                continue

            region_type = None

            if det_type in ["face", "person", "object", "subject"]:
                region_type = "PRIMARY_FOCAL_REGION"
            elif det_type in ["text", "ocr"]:
                region_type = "TEXT_REGION"
            elif det_type == "logo":
                region_type = "LOGO_REGION"

            if region_type:
                regions.append({
                    "type": region_type,
                    "box": tuple(map(int, box)),
                    "score": float(det.get("score", 85)),
                    "source": "hf_optional_ai",
                })

    return sorted(regions, key=lambda r: r["score"], reverse=True)

def cover_resize_no_padding(img, target_w, target_h):
    src_w, src_h = img.size
    scale = max(target_w / src_w, target_h / src_h)
    new_w, new_h = int(src_w * scale), int(src_h * scale)

    resized = img.resize((new_w, new_h), Image.LANCZOS)

    left = max(0, (new_w - target_w) // 2)
    top = max(0, (new_h - target_h) // 2)

    return resized.crop((left, top, left + target_w, top + target_h))

def contain_resize_with_transparency(img, target_w, target_h):
    src_w, src_h = img.size
    scale = min(target_w / src_w, target_h / src_h)
    new_w, new_h = int(src_w * scale), int(src_h * scale)

    resized = img.resize((new_w, new_h), Image.LANCZOS)

    canvas = Image.new("RGBA", (target_w, target_h), (0, 0, 0, 0))
    canvas.paste(resized, ((target_w - new_w) // 2, (target_h - new_h) // 2))

    return canvas

def tile_texture(img, target_w, target_h):
    tile_size = max(300, min(900, target_w, target_h))
    tile = cover_resize_no_padding(img, tile_size, tile_size)

    out = Image.new("RGBA", (target_w, target_h))

    for x in range(0, target_w, tile.width):
        for y in range(0, target_h, tile.height):
            out.paste(tile, (x, y))

    return out.crop((0, 0, target_w, target_h))

def get_mask_path(template_id, piece_name):
    candidates = [
        f"template_assets/{template_id}/masks/{piece_name}.png",
        f"template_assets/{template_id}/masks/{norm(piece_name)}.png",
        f"template_assets/{template_id}/{piece_name}.png",
        f"template_assets/{template_id}/{norm(piece_name)}.png",
    ]

    for path in candidates:
        if os.path.exists(path):
            return path

    return None

def apply_piece_mask(mapped_img, template_id, piece_name):
    mask_path = get_mask_path(template_id, piece_name)

    if not mask_path:
        return mapped_img, None

    mask_img = Image.open(mask_path).convert("RGBA").resize(mapped_img.size, Image.LANCZOS)
    alpha = mask_img.getchannel("A")

    out = mapped_img.convert("RGBA")
    out.putalpha(alpha)

    return out, mask_path

def choose_region(piece_name, regions):
    rule = AOP_HOODIE_MAPPING_RULES.get(norm(piece_name))

    if not rule:
        return regions[0], "safe_cover"

    for preferred in rule["preferred"]:
        for region in regions:
            if region["type"] == preferred and region["type"] not in rule["forbidden"]:
                return region, rule["strategy"]

    for region in regions:
        if region["type"] not in rule["forbidden"]:
            return region, rule["strategy"]

    env = next((r for r in regions if r["type"] == "ENVIRONMENT_REGION"), regions[0])
    return env, "fallback_texture"

def apply_strategy(image, region, strategy, target_w, target_h):
    crop = image.crop(region["box"]).convert("RGBA")

    if strategy == "texture_band":
        return tile_texture(crop, target_w, target_h)

    if strategy == "vertical_texture":
        return cover_resize_no_padding(crop, target_w, target_h)

    if strategy == "dark_texture":
        return cover_resize_no_padding(crop, target_w, target_h)

    if strategy == "fallback_texture":
        return cover_resize_no_padding(crop, target_w, target_h)

    if strategy == "centered_focal":
        return cover_resize_no_padding(crop, target_w, target_h)

    if strategy == "wide_dramatic":
        return cover_resize_no_padding(crop, target_w, target_h)

    if strategy == "lower_anchor":
        return cover_resize_no_padding(crop, target_w, target_h)

    return cover_resize_no_padding(crop, target_w, target_h)

def map_all(image_path, template_id, pieces, prompt, use_optional_ai=False):
    image = Image.open(image_path).convert("RGBA")

    ai_results = run_optional_ai_processors(image_path, prompt) if use_optional_ai else {}
    regions = analyze_regions(image, prompt, ai_results)

    outputs = {}
    decisions = []
    warnings = []
    used_boxes = {}
    mask_usage = {}

    for piece_name, size in pieces.items():
        target_w, target_h = size

        region, strategy = choose_region(piece_name, regions)
        mapped = apply_strategy(image, region, strategy, target_w, target_h)
        mapped, mask_path = apply_piece_mask(mapped, template_id, piece_name)

        outputs[piece_name] = mapped

        decision = {
            "piece": piece_name,
            "region": region["type"],
            "strategy": strategy,
            "box": region["box"],
            "score": region["score"],
            "output_size": [target_w, target_h],
            "mask_applied": bool(mask_path),
            "mask_path": mask_path,
        }

        decisions.append(decision)

        key = str(region["box"])
        used_boxes[key] = used_boxes.get(key, 0) + 1
        mask_usage[piece_name] = bool(mask_path)

        risky_small = norm(piece_name) in [
            "left_sleeve",
            "right_sleeve",
            "collar",
            "left_cuff",
            "right_cuff",
            "waistband",
            "kangaroo_pocket",
        ]

        if risky_small and region["type"] in [
            "TEXT_REGION",
            "LOGO_REGION",
            "PRIMARY_FOCAL_REGION",
        ]:
            warnings.append(
                f"{piece_name}: readable text, logo, or primary focal content may be on a risky small panel."
            )

        if not mask_path:
            warnings.append(
                f"{piece_name}: no piece mask found, output is rectangular. Add template_assets/{template_id}/masks/{piece_name}.png for true shaped AOP output."
            )

    front_decisions = [
        d for d in decisions if norm(d["piece"]) in ["front", "front_body"]
    ]

    if front_decisions and front_decisions[0]["region"] != "PRIMARY_FOCAL_REGION":
        warnings.append("Main focal point was not placed on the front panel.")

    if used_boxes and max(used_boxes.values()) > 3:
        warnings.append("Too many panels are using the same crop. Design may look repetitive.")

    if not any(mask_usage.values()):
        warnings.append(
            "No template masks were applied. Current outputs are rectangular crops, not true garment-shaped production pieces."
        )

    return outputs, {
        "prompt": prompt,
        "template_id": template_id,
        "plain_engine_logic": [
            "Big panels get meaning.",
            "Small panels get texture.",
            "Risky seams get atmosphere.",
            "Pockets get lower anchors.",
            "Cuffs get color bands.",
            "Hoods get continuation, not the whole story.",
            "True AOP piece shape requires alpha masks from template PNGs.",
        ],
        "regions": regions,
        "decisions": decisions,
        "warnings": warnings,
        "optional_ai_processors": ai_results,
    }