File size: 13,815 Bytes
a49f337
93af3e2
b7cfbcf
a49f337
c1ad781
597b21e
 
93af3e2
74168bc
93af3e2
 
 
 
 
58d1893
74168bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
597b21e
93af3e2
 
 
 
 
 
 
 
 
 
597b21e
93af3e2
 
 
a49f337
93af3e2
 
 
 
 
 
 
 
 
a49f337
 
93af3e2
 
a49f337
 
93af3e2
b0a0a29
 
93af3e2
 
b0a0a29
 
93af3e2
a49f337
 
 
 
 
 
 
 
 
93af3e2
a49f337
 
 
 
 
 
 
 
 
74168bc
 
a49f337
 
 
 
 
74168bc
 
 
 
 
 
 
 
 
 
a49f337
b0a0a29
a49f337
b0a0a29
a49f337
 
b0a9f3e
 
 
a49f337
b0a9f3e
93af3e2
58d1893
a49f337
 
 
 
 
 
 
 
 
93af3e2
 
a49f337
 
 
b0a9f3e
a49f337
 
 
b0a9f3e
a49f337
b0a9f3e
93af3e2
9047ade
 
 
 
 
 
 
 
 
c2d1871
9047ade
74168bc
9047ade
74168bc
 
a49f337
 
 
74168bc
 
 
 
 
 
9047ade
 
 
 
 
 
c2d1871
9047ade
 
93af3e2
 
9047ade
93af3e2
a49f337
 
93af3e2
 
 
 
9047ade
93af3e2
 
a49f337
 
 
 
 
 
 
 
b0a9f3e
a49f337
b0a9f3e
 
 
a49f337
 
b0a9f3e
a49f337
 
 
b0a9f3e
9047ade
 
74168bc
 
 
9047ade
74168bc
 
 
 
b0a9f3e
a49f337
 
 
 
b0a9f3e
a49f337
 
 
 
 
b0a9f3e
a49f337
 
b0a9f3e
a49f337
 
b0a9f3e
a49f337
 
b0a0a29
a49f337
 
 
 
 
 
3bb8a2e
a49f337
 
 
 
b0a9f3e
a49f337
 
 
b0a9f3e
a49f337
b0a9f3e
 
b0a0a29
b0a9f3e
a49f337
 
 
 
 
b0a9f3e
 
a49f337
b0a9f3e
a49f337
b0a9f3e
b0a0a29
a49f337
93af3e2
 
a49f337
b0a0a29
a49f337
93af3e2
9047ade
 
 
93af3e2
b0a9f3e
b0a0a29
93af3e2
 
a49f337
 
 
 
 
93af3e2
a1ef78c
a49f337
9047ade
a49f337
 
93af3e2
 
a1ef78c
9047ade
 
 
 
 
 
 
c2d1871
9047ade
 
 
 
 
 
 
3bb8a2e
9047ade
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a49f337
9047ade
 
 
93af3e2
 
a49f337
9047ade
a49f337
 
 
b0a0a29
b0a9f3e
a49f337
93af3e2
 
a49f337
 
 
 
 
 
 
b0a9f3e
a49f337
93af3e2
b0a0a29
a49f337
93af3e2
 
 
a49f337
 
 
 
 
 
9047ade
93af3e2
a49f337
93af3e2
a1ef78c
93af3e2
a49f337
9047ade
a49f337
9047ade
93af3e2
 
c1ad781
93af3e2
b0a0a29
a49f337
b0a0a29
 
 
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
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
import os
import random
import warnings
import gc
import gradio as gr
import numpy as np
import spaces
import torch
import torch.nn as nn
from diffusers import FluxImg2ImgPipeline
from gradio_imageslider import ImageSlider
from PIL import Image
from huggingface_hub import snapshot_download
import requests

# Minimal ESRGAN implementation (without basicsr dependency)
class ResidualDenseBlock(nn.Module):
    def __init__(self, num_feat=64, num_grow_ch=32):
        super(ResidualDenseBlock, self).__init__()
        self.conv1 = nn.Conv2d(num_feat, num_grow_ch, 3, 1, 1)
        self.conv2 = nn.Conv2d(num_feat + num_grow_ch, num_grow_ch, 3, 1, 1)
        self.conv3 = nn.Conv2d(num_feat + 2 * num_grow_ch, num_grow_ch, 3, 1, 1)
        self.conv4 = nn.Conv2d(num_feat + 3 * num_grow_ch, num_grow_ch, 3, 1, 1)
        self.conv5 = nn.Conv2d(num_feat + 4 * num_grow_ch, num_feat, 3, 1, 1)
        self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True)

    def forward(self, x):
        x1 = self.lrelu(self.conv1(x))
        x2 = self.lrelu(self.conv2(torch.cat((x, x1), 1)))
        x3 = self.lrelu(self.conv3(torch.cat((x, x1, x2), 1)))
        x4 = self.lrelu(self.conv4(torch.cat((x, x1, x2, x3), 1)))
        x5 = self.conv5(torch.cat((x, x1, x2, x3, x4), 1))
        return x5 * 0.2 + x

class RRDB(nn.Module):
    def __init__(self, num_feat, num_grow_ch=32):
        super(RRDB, self).__init__()
        self.rdb1 = ResidualDenseBlock(num_feat, num_grow_ch)
        self.rdb2 = ResidualDenseBlock(num_feat, num_grow_ch)
        self.rdb3 = ResidualDenseBlock(num_feat, num_grow_ch)

    def forward(self, x):
        out = self.rdb1(x)
        out = self.rdb2(out)
        out = self.rdb3(out)
        return out * 0.2 + x

class RRDBNet(nn.Module):
    def __init__(self, num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23, num_grow_ch=32, scale=4):
        super(RRDBNet, self).__init__()
        self.scale = scale
        
        self.conv_first = nn.Conv2d(num_in_ch, num_feat, 3, 1, 1)
        self.body = nn.Sequential(*[RRDB(num_feat, num_grow_ch) for _ in range(num_block)])
        self.conv_body = nn.Conv2d(num_feat, num_feat, 3, 1, 1)
        
        # Upsampling
        self.conv_up1 = nn.Conv2d(num_feat, num_feat, 3, 1, 1)
        self.conv_up2 = nn.Conv2d(num_feat, num_feat, 3, 1, 1)
        self.conv_hr = nn.Conv2d(num_feat, num_feat, 3, 1, 1)
        self.conv_last = nn.Conv2d(num_feat, num_out_ch, 3, 1, 1)
        self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True)

    def forward(self, x):
        fea = self.conv_first(x)
        trunk = self.conv_body(self.body(fea))
        fea = fea + trunk

        fea = self.lrelu(self.conv_up1(nn.functional.interpolate(fea, scale_factor=2, mode='nearest')))
        fea = self.lrelu(self.conv_up2(nn.functional.interpolate(fea, scale_factor=2, mode='nearest')))
        out = self.conv_last(self.lrelu(self.conv_hr(fea)))
        return out

css = """
#col-container {
    margin: 0 auto;
    max-width: 800px;
}
.main-header {
    text-align: center;
    margin-bottom: 2rem;
}
"""

# Get HuggingFace token
huggingface_token = os.getenv("HF_TOKEN")

# Download FLUX model if not already cached
print("πŸ“₯ Downloading FLUX model...")
model_path = snapshot_download(
    repo_id="black-forest-labs/FLUX.1-dev", 
    repo_type="model", 
    ignore_patterns=["*.md", "*.gitattributes"],
    local_dir="FLUX.1-dev",
    token=huggingface_token,
)

# Load FLUX pipeline on CPU initially
print("πŸ“₯ Loading FLUX Img2Img pipeline...")
pipe = FluxImg2ImgPipeline.from_pretrained(
    model_path, 
    torch_dtype=torch.bfloat16,
    use_safetensors=True
)

# Enable memory optimizations
pipe.enable_vae_tiling()
pipe.enable_vae_slicing()
pipe.vae.enable_tiling()
pipe.vae.enable_slicing()

# Download and load ESRGAN 4x-UltraSharp model
print("πŸ“₯ Loading ESRGAN 4x-UltraSharp...")
esrgan_path = "4x-UltraSharp.pth"
if not os.path.exists(esrgan_path):
    print("Downloading ESRGAN model...")
    url = "https://huggingface.co/uwg/upscaler/resolve/main/ESRGAN/4x-UltraSharp.pth"
    response = requests.get(url)
    with open(esrgan_path, "wb") as f:
        f.write(response.content)

# Initialize ESRGAN model
esrgan_model = RRDBNet(
    num_in_ch=3, 
    num_out_ch=3, 
    num_feat=64, 
    num_block=23, 
    num_grow_ch=32, 
    scale=4
)

# Load state dict
state_dict = torch.load(esrgan_path, map_location='cpu')
if 'params_ema' in state_dict:
    state_dict = state_dict['params_ema']
elif 'params' in state_dict:
    state_dict = state_dict['params']

# Clean state dict keys if needed
cleaned_state_dict = {}
for k, v in state_dict.items():
    if k.startswith('module.'):
        cleaned_state_dict[k[7:]] = v
    else:
        cleaned_state_dict[k] = v

esrgan_model.load_state_dict(cleaned_state_dict, strict=False)
esrgan_model.eval()

print("βœ… All models loaded successfully!")

MAX_SEED = 1000000
MAX_INPUT_SIZE = 512  # Max input size before upscaling


def make_multiple_16(n):
    """Round to nearest multiple of 16 for FLUX requirements"""
    return ((n + 15) // 16) * 16


def truncate_prompt(prompt, max_tokens=75):
    """Truncate prompt to avoid CLIP token limit (77 tokens)"""
    if not prompt:
        return ""
    
    # Simple truncation by character count (rough approximation)
    if len(prompt) > 250:  # ~75 tokens
        return prompt[:250] + "..."
    return prompt


def prepare_image(image, max_size=MAX_INPUT_SIZE):
    """Prepare image for processing"""
    w, h = image.size
    
    # Limit input size
    if w > max_size or h > max_size:
        image.thumbnail((max_size, max_size), Image.LANCZOS)
    
    return image


def esrgan_upscale(image, model, device='cuda', upscale_factor=4):
    """Upscale image using ESRGAN with variable factor support"""
    orig_w, orig_h = image.size
    pre_resize_factor = upscale_factor / 4.0
    low_res_w = int(orig_w * pre_resize_factor)
    low_res_h = int(orig_h * pre_resize_factor)
    if low_res_w < 1 or low_res_h < 1:
        raise ValueError("Upscale factor too small for image size")
    
    low_res_image = image.resize((low_res_w, low_res_h), Image.BICUBIC)  # Changed to BICUBIC for better match to training degradation
    
    # Prepare image
    img_np = np.array(low_res_image).astype(np.float32) / 255.
    img_np = np.transpose(img_np, (2, 0, 1))  # HWC to CHW
    img_tensor = torch.from_numpy(img_np).unsqueeze(0).to(device)
    
    # Upscale
    with torch.no_grad():
        output = model(img_tensor)
        output = output.squeeze(0).cpu().clamp(0, 1)
        output_np = output.numpy()
        output_np = np.transpose(output_np, (1, 2, 0))  # CHW to HWC
        output_np = (output_np * 255).astype(np.uint8)
    
    upscaled = Image.fromarray(output_np)
    
    # Resize back to exact target size if needed (due to rounding)
    target_w = int(orig_w * upscale_factor)
    target_h = int(orig_h * upscale_factor)
    if upscaled.size != (target_w, target_h):
        upscaled = upscaled.resize((target_w, target_h), Image.BICUBIC)  # Changed to BICUBIC
    
    return upscaled


@spaces.GPU(duration=120)  # Increased to 120 seconds
def enhance_image(
    input_image,
    prompt,
    seed,
    randomize_seed,
    num_inference_steps,
    denoising_strength,
    upscale_factor,
    progress=gr.Progress(track_tqdm=True),
):
    """Main enhancement function"""
    if input_image is None:
        raise gr.Error("Please upload an image")
    
    # Clear memory
    torch.cuda.empty_cache()
    gc.collect()
    
    try:
        # Randomize seed if needed
        if randomize_seed:
            seed = random.randint(0, MAX_SEED)
        
        # Prepare and validate prompt
        prompt = truncate_prompt(prompt.strip() if prompt else "high quality, detailed")
        
        # Prepare input image
        input_image = prepare_image(input_image)
        original_size = input_image.size
        
        # Step 1: ESRGAN upscale on GPU
        gr.Info(f"πŸ” Upscaling with ESRGAN x{upscale_factor}...")
        
        # Move ESRGAN to GPU for faster processing
        esrgan_model.to("cuda")
        upscaled_image = esrgan_upscale(input_image, esrgan_model, device="cuda", upscale_factor=upscale_factor)
        
        # Move ESRGAN back to CPU to free memory
        esrgan_model.to("cpu")
        torch.cuda.empty_cache()
        
        # Ensure dimensions are multiples of 16 for FLUX
        w, h = upscaled_image.size
        new_w = make_multiple_16(w)
        new_h = make_multiple_16(h)
        
        if new_w != w or new_h != h:
            # Pad image to meet requirements
            padded = Image.new('RGB', (new_w, new_h))
            padded.paste(upscaled_image, (0, 0))
            upscaled_image = padded
        
        # Step 2: FLUX enhancement
        gr.Info("🎨 Enhancing with FLUX...")
        
        # Move pipeline to GPU
        pipe.to("cuda")
        
        # Generate with FLUX
        generator = torch.Generator(device="cuda").manual_seed(seed)
        
        with torch.inference_mode():
            result = pipe(
                prompt=prompt,
                image=upscaled_image,
                strength=denoising_strength,
                num_inference_steps=num_inference_steps,
                guidance_scale=3.5,  # Recommended for FLUX.1-dev to reduce artifacts
                height=new_h,
                width=new_w,
                generator=generator,
            ).images[0]
        
        # Crop back if we padded
        if new_w != w or new_h != h:
            result = result.crop((0, 0, w, h))
        
        # Move pipeline back to CPU
        pipe.to("cpu")
        torch.cuda.empty_cache()
        gc.collect()
        
        # Prepare images for slider (before/after)
        input_resized = input_image.resize(result.size, Image.LANCZOS)
        
        gr.Info("βœ… Enhancement complete!")
        return [input_resized, result], seed
        
    except Exception as e:
        # Cleanup on error
        pipe.to("cpu")
        esrgan_model.to("cpu")
        torch.cuda.empty_cache()
        gc.collect()
        raise gr.Error(f"Enhancement failed: {str(e)}")


# Create Gradio interface
with gr.Blocks(css=css) as demo:
    gr.HTML("""
    <div class="main-header">
        <h1>πŸš€ Flux Dev Ultimate Upscaler</h1>
        <p>Upload an image to upscale 2-4x with ESRGAN and enhance with FLUX</p>
        <p>Optimized for <strong>ZeroGPU</strong> | Max input: 512x512 β†’ Output: up to 2048x2048</p>
    </div>
    """)
    
    with gr.Row():
        with gr.Column(scale=1):
            # Input section
            input_image = gr.Image(
                label="Input Image",
                type="pil",
                height=256
            )
            
            prompt = gr.Textbox(
                label="Describe image with prompt",
                placeholder="Describe the desired enhancement (e.g., 'high quality, sharp details, vibrant colors')",
                value="high quality, ultra detailed, sharp",
                lines=2
            )
            
            # Advanced Settings (always open)
            upscale_factor = gr.Slider(
                label="Upscale Ratio",
                minimum=2,
                maximum=4,
                step=1,
                value=4,
                info="Choose upscale factor (2x, 3x, 4x). Use 4x for best results; lower may cause color artifacts."
            )
            
            num_inference_steps = gr.Slider(
                label="Enhancement Steps",
                minimum=10,
                maximum=25,
                step=1,
                value=20,  # Increased default for better denoising
                info="More steps = better quality but slower"
            )
            
            denoising_strength = gr.Slider(
                label="Creativity (Denoising)",
                minimum=0.1,
                maximum=0.6,
                step=0.05,
                value=0.35,
                info="Higher = more changes to the image"
            )
            
            with gr.Row():
                randomize_seed = gr.Checkbox(
                    label="Randomize seed",
                    value=True
                )
                seed = gr.Number(
                    label="Seed",
                    value=42
                )
            
            enhance_btn = gr.Button(
                "Upscale",
                variant="primary",
                size="lg"
            )
        
        with gr.Column(scale=2):
            # Output section
            result_slider = ImageSlider(
                type="pil",
                label="Before / After",
                interactive=False,
                height=512
            )
            
            used_seed = gr.Number(
                label="Seed Used",
                interactive=False,
                visible=False
            )
    
    # Event handler
    enhance_btn.click(
        fn=enhance_image,
        inputs=[
            input_image,
            prompt,
            seed,
            randomize_seed,
            num_inference_steps,
            denoising_strength,
            upscale_factor,
        ],
        outputs=[result_slider, used_seed]
    )
    
    gr.HTML("""
    <div style="margin-top: 2rem; text-align: center; color: #666;">
        <p>πŸ“Œ Pipeline: ESRGAN 2-4x-UltraSharp β†’ FLUX Dev Enhancement</p>
        <p>⚑ Optimized for ZeroGPU with automatic memory management</p>
        <p>πŸ“Œ Note: User is responsible for obtaining commercial license from Flux Dev if using image commercially under their license.</p>
    </div>
    """)

if __name__ == "__main__":
    demo.queue(max_size=3).launch(
        share=False,
        server_name="0.0.0.0",
        server_port=7860
    )