Upload DiffSketcher/Examples.md with huggingface_hub
Browse files- DiffSketcher/Examples.md +301 -0
DiffSketcher/Examples.md
ADDED
@@ -0,0 +1,301 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Qualitative Results
|
2 |
+
|
3 |
+
### Case: Horse Drinking Water
|
4 |
+
|
5 |
+
**JVSP + ASDS fine-tuning (horse)**
|
6 |
+
|
7 |
+
**Preview:**
|
8 |
+
|
9 |
+
| canvas size / Rendering | Strokes Initialization | 100 step | visual best |
|
10 |
+
|-------------------------|---------------------------------------------------------|-----------------------------------------------------------|:---------------------------------------------------------:|
|
11 |
+
| `canvas_size=224` | <img src="./img/HorseDrinkingWater/svg_iter0_C224.svg"> | <img src="./img/HorseDrinkingWater/svg_iter100_C224.svg"> | <img src="./img/HorseDrinkingWater/visual_best_C224.svg"> |
|
12 |
+
| `canvas_size=600` | <img src="./img/HorseDrinkingWater/svg_iter0_C600.svg"> | <img src="./img/HorseDrinkingWater/svg_iter100_C600.svg"> | <img src="./img/HorseDrinkingWater/visual_best_C600.svg"> |
|
13 |
+
|
14 |
+
**Script:**
|
15 |
+
|
16 |
+
```shell
|
17 |
+
# canvas_size: 224
|
18 |
+
python run_painterly_render.py -c diffsketcher.yaml -eval_step 10 -save_step 10 -update "token_ind=2 num_paths=96" -pt "A horse is drinking water by the lake" -respath ./workdir/draw_horse -d 998
|
19 |
+
# canvas_size: 600
|
20 |
+
python run_painterly_render.py -c diffsketcher.yaml -eval_step 10 -save_step 10 -update "token_ind=2 num_paths=96 image_size=600 width=3.5" -pt "A horse is drinking water by the lake" -respath ./workdir/draw_horse -d 998
|
21 |
+
```
|
22 |
+
|
23 |
+
**train from scratch via ASDS loss + SDXL (horse)**
|
24 |
+
|
25 |
+
```shell
|
26 |
+
python run_painterly_render.py -c diffsketcher.yaml -eval_step 10 -save_step 10 -update "image_size=1110 token_ind=2 num_paths=96 sds.grad_scale=2 sds.warmup=0 sds.crop_size=1024 clip.vis_loss=0 perceptual.coeff=0 opacity_delta=0.2 num_iter=2000 model_id=sdxl" -pt "A horse is drinking water by the lake" -respath ./workdir/draw_horse -d 998 --download
|
27 |
+
```
|
28 |
+
|
29 |
+
**JVSP + ASDS fine-tune (horse) + including width**
|
30 |
+
|
31 |
+
**Preview:**
|
32 |
+
|
33 |
+
| <img src="./img/HorseDrinkingWater-ink/svg_iter0.svg"> | <img src="./img/HorseDrinkingWater-ink/svg_iter100.svg"> | <img src="./img/HorseDrinkingWater-ink/svg_iter1000.svg"> |
|
34 |
+
|--------------------------------------------------------|----------------------------------------------------------|-----------------------------------------------------------|
|
35 |
+
| Strokes Initialization | 100 step | 1000 step |
|
36 |
+
|
37 |
+
**Script:**
|
38 |
+
|
39 |
+
```shell
|
40 |
+
python run_painterly_render.py -c diffsketcher-width.yaml -eval_step 10 -save_step 10 -update "token_ind=2 num_paths=96 num_iter=1000 grad_scale=0" -pt "A horse is drinking water by the lake" -respath ./workdir/draw_horse_ink -d 998
|
41 |
+
```
|
42 |
+
|
43 |
+
### Case: 3D Style Sketch
|
44 |
+
|
45 |
+
**Preview:**
|
46 |
+
|
47 |
+
| <img src="./img/3D_rose/svg_iter0.svg"> | <img src="./img/3D_rose/svg_iter100.svg"> | <img src="./img/3D_rose/visual_best.svg"> |
|
48 |
+
|-----------------------------------------|-------------------------------------------|-------------------------------------------|
|
49 |
+
| Strokes Initialization | 100 step | 1510 step |
|
50 |
+
|
51 |
+
**Script:**
|
52 |
+
|
53 |
+
```shell
|
54 |
+
python run_painterly_render.py -c diffsketcher.yaml -eval_step 10 -save_step 10 -update "token_ind=4 num_paths=128 sds.grad_scale=0" -pt "A 3d single rose" -respath ./workdir/3d_rose -d 291516
|
55 |
+
```
|
56 |
+
|
57 |
+
### Case: Elephant
|
58 |
+
|
59 |
+
**Preview:**
|
60 |
+
|
61 |
+
| <img src="./img/elephant/svg_iter0.svg"> | <img src="./img/elephant/svg_iter100.svg"> | <img src="./img/elephant/semantic_best.svg"> |
|
62 |
+
|------------------------------------------|--------------------------------------------|----------------------------------------------|
|
63 |
+
| Strokes Initialization | 100 step | 1890 step |
|
64 |
+
|
65 |
+
**Script:**
|
66 |
+
|
67 |
+
```shell
|
68 |
+
python run_painterly_render.py -c diffsketcher.yaml -eval_step 10 -save_step 10 -update "token_ind=2 num_paths=128 softmax_temp=0.4 sds.grad_scale=1 sds.warmup=0 clip.vis_loss=0 perceptual.coeff=0 lr_scheduler=True num_iter=2000 opacity_delta=0.1" -pt "an elephant. minimal 2d line drawing. trending on artstation." -respath ./workdir/elephant -d 197920
|
69 |
+
```
|
70 |
+
|
71 |
+
### Case: Yoda
|
72 |
+
|
73 |
+
**Preview:**
|
74 |
+
|
75 |
+
| <img src="./img/Yoda/svg_iter0.svg"> | <img src="./img/Yoda/svg_iter100.svg"> | <img src="./img/Yoda/visual_best.svg"> |
|
76 |
+
|--------------------------------------|----------------------------------------|----------------------------------------|
|
77 |
+
| Strokes Initialization | 100 step | 1780 step |
|
78 |
+
|
79 |
+
**Script:**
|
80 |
+
|
81 |
+
```shell
|
82 |
+
CUDA_VISIBLE_DEVICES=0 python run_painterly_render.py \
|
83 |
+
-c diffsketcher.yaml \
|
84 |
+
-eval_step 10 -save_step 10 \
|
85 |
+
-update "token_ind=7 num_paths=96 sds.warmup=1500 num_iter=2000" \
|
86 |
+
-pt "Very detailed masterpiece painting of baby yoda hoding a lightsaber, portrait, artstation, concept art by greg rutkowski" \
|
87 |
+
-respath ./workdir/Yoda \
|
88 |
+
-d 998
|
89 |
+
```
|
90 |
+
|
91 |
+
### Case: Fox
|
92 |
+
|
93 |
+
**Preview:**
|
94 |
+
|
95 |
+
| <img src="./img/Fox/svg_iter0.svg"> | <img src="./img/Fox/svg_iter100.svg"> | <img src="./img/Fox/visual_best.svg"> |
|
96 |
+
|-------------------------------------|---------------------------------------|---------------------------------------|
|
97 |
+
| Strokes Initialization | 100 step | 1500 step |
|
98 |
+
|
99 |
+
**Script:**
|
100 |
+
|
101 |
+
```shell
|
102 |
+
CUDA_VISIBLE_DEVICES=0 python run_painterly_render.py \
|
103 |
+
-c diffsketcher.yaml \
|
104 |
+
-eval_step 10 -save_step 10 \
|
105 |
+
-update "token_ind=2 num_paths=96 softmax_temp=0.5 num_iter=2000" \
|
106 |
+
-pt "A fox is sitting on the sofa" \
|
107 |
+
-respath ./workdir/fox \
|
108 |
+
-d 9007
|
109 |
+
```
|
110 |
+
|
111 |
+
### Case: Balloons
|
112 |
+
|
113 |
+
**Preview:**
|
114 |
+
|
115 |
+
| <img src="./img/balloons/svg_iter0.svg"> | <img src="./img/balloons/svg_iter100.svg"> | <img src="./img/balloons/visual_best.svg"> |
|
116 |
+
|------------------------------------------|--------------------------------------------|--------------------------------------------|
|
117 |
+
| Strokes Initialization | 100 step | 2000 step |
|
118 |
+
|
119 |
+
**Script:**
|
120 |
+
|
121 |
+
```shell
|
122 |
+
CUDA_VISIBLE_DEVICES=0 python run_painterly_render.py \
|
123 |
+
-c diffsketcher.yaml \
|
124 |
+
-eval_step 10 -save_step 10 \
|
125 |
+
-update "token_ind=4 num_paths=128 softmax_temp=0.5" \
|
126 |
+
-pt "Colorful hot air balloons high over the mountains" \
|
127 |
+
-respath ./workdir/balloons \
|
128 |
+
-d 9998
|
129 |
+
```
|
130 |
+
|
131 |
+
### Case: A cat on a bicycle
|
132 |
+
|
133 |
+
**Preview:**
|
134 |
+
|
135 |
+
| <img src="./img/cat_ride_bike/svg_iter0.svg"> | <img src="./img/cat_ride_bike/svg_iter100.svg"> | <img src="./img/cat_ride_bike/visual_best.svg"> |
|
136 |
+
|-----------------------------------------------|-------------------------------------------------|-------------------------------------------------|
|
137 |
+
| Strokes Initialization | 100 step | 1800 step |
|
138 |
+
|
139 |
+
**Script:**
|
140 |
+
|
141 |
+
```shell
|
142 |
+
CUDA_VISIBLE_DEVICES=0 python run_painterly_render.py \
|
143 |
+
-c diffsketcher.yaml \
|
144 |
+
-eval_step 10 -save_step 10 \
|
145 |
+
-update "token_ind=2 num_paths=48 ssoftmax_temp=0.5 mask_object=False" \
|
146 |
+
-pt "A cute cat in the style of Pixar animations, wearing a helmet and riding a bike" \
|
147 |
+
-respath ./workdir/cat \
|
148 |
+
-d 8030
|
149 |
+
```
|
150 |
+
|
151 |
+
### Case: Cat
|
152 |
+
|
153 |
+
**Preview:**
|
154 |
+
|
155 |
+
| <img src="./img/cat/svg_iter0.svg"> | <img src="./img/cat/svg_iter100.svg"> | <img src="./img/cat/final_svg.svg"> |
|
156 |
+
|-------------------------------------|---------------------------------------|-------------------------------------|
|
157 |
+
| Strokes Initialization | 100 step | 2000 step |
|
158 |
+
|
159 |
+
**Script:**
|
160 |
+
|
161 |
+
```shell
|
162 |
+
python run_painterly_render.py -c diffsketcher.yaml -eval_step 10 -save_step 10 -update "token_ind=2 num_paths=128 softmax_temp=0.4 sds.num_aug=4 sds.grad_scale=1 sds.warmup=0 clip.vis_loss=0 perceptual.coeff=0 lr_scheduler=True num_iter=2000 opacity_delta=0.3" -pt "A cat. minimal 2d line drawing. trending on artstation." -respath ./workdir/cat -d 915346
|
163 |
+
```
|
164 |
+
|
165 |
+
### Case: Athens
|
166 |
+
|
167 |
+
**Preview:**
|
168 |
+
|
169 |
+
| <img src="./img/Athens/svg_iter0.svg"> | <img src="./img/Athens/svg_iter100.svg"> | <img src="./img/Athens/visual_best.svg"> |
|
170 |
+
|----------------------------------------|------------------------------------------|------------------------------------------|
|
171 |
+
| Strokes Initialization | 100 step | 1380 step |
|
172 |
+
|
173 |
+
**Script:**
|
174 |
+
|
175 |
+
```shell
|
176 |
+
python run_painterly_render.py \
|
177 |
+
-c diffsketcher.yaml \
|
178 |
+
-eval_step 10 -save_step 10 \
|
179 |
+
-update "token_ind=11 num_paths=140 softmax_temp=0.5 num_iter=2000" \
|
180 |
+
-pt "A loose ink sketching with watercolors of a modern Athens neighborhood, architectural, detailed, old building and new buildings, quiet street" \
|
181 |
+
-respath ./workdir/neighborhood \
|
182 |
+
-d 42
|
183 |
+
```
|
184 |
+
|
185 |
+
### Case: Macaw
|
186 |
+
|
187 |
+
**Preview:**
|
188 |
+
|
189 |
+
| <img src="./img/Macaw/svg_iter0.svg"> | <img src="./img/Macaw/svg_iter100.svg"> | <img src="./img/Macaw/visual_best.svg"> |
|
190 |
+
|---------------------------------------|-----------------------------------------|-----------------------------------------|
|
191 |
+
| Strokes Initialization | 100 step | 1200 step |
|
192 |
+
|
193 |
+
**Script:**
|
194 |
+
|
195 |
+
```shell
|
196 |
+
CUDA_VISIBLE_DEVICES=0 python run_painterly_render.py \
|
197 |
+
-c diffsketcher.yaml \
|
198 |
+
-eval_step 10 -save_step 10 \
|
199 |
+
-update "token_ind=1 num_paths=96 sds.warmup=1500 num_iter=2000" \
|
200 |
+
-pt "macaw full color, ultra detailed, realistic, insanely beautiful" \
|
201 |
+
-respath ./workdir/macaw \
|
202 |
+
-d 8091
|
203 |
+
```
|
204 |
+
|
205 |
+
### Case: Bunny
|
206 |
+
|
207 |
+
**Preview:**
|
208 |
+
|
209 |
+
| <img src="./img/Bunny/svg_iter0.svg"> | <img src="./img/Bunny/svg_iter100.svg"> | <img src="./img/Bunny/visual_best.svg"> |
|
210 |
+
|---------------------------------------|-----------------------------------------|-----------------------------------------|
|
211 |
+
| Strokes Initialization | 100 step | 670 step |
|
212 |
+
|
213 |
+
**Script:**
|
214 |
+
|
215 |
+
```shell
|
216 |
+
CUDA_VISIBLE_DEVICES=0 python run_painterly_render.py \
|
217 |
+
-c diffsketcher.yaml \
|
218 |
+
-eval_step 10 -save_step 10 \
|
219 |
+
-update "token_ind=5 num_paths=96 softmax_temp=0.5 num_iter=2000" \
|
220 |
+
-pt "portrait of two white bunnies, super realistic, highly detailed" \
|
221 |
+
-respath ./workdir/bunny \
|
222 |
+
-d 9001
|
223 |
+
```
|
224 |
+
|
225 |
+
### Case: Dragon
|
226 |
+
|
227 |
+
**Preview:**
|
228 |
+
|
229 |
+
| <img src="./img/Dragon/svg_iter0.svg"> | <img src="./img/Dragon/svg_iter100.svg"> | <img src="./img/Dragon/visual_best.svg"> |
|
230 |
+
|----------------------------------------|------------------------------------------|------------------------------------------|
|
231 |
+
| Strokes Initialization | 100 step | 1200 step |
|
232 |
+
|
233 |
+
**Script:**
|
234 |
+
|
235 |
+
```shell
|
236 |
+
CUDA_VISIBLE_DEVICES=0 python run_painterly_render.py \
|
237 |
+
-c diffsketcher.yaml \
|
238 |
+
-eval_step 10 -save_step 10 \
|
239 |
+
-update "token_ind=2 num_paths=64 softmax_temp=0.5 mask_object=True" \
|
240 |
+
-pt "A dragon flying in the sky, full body" \
|
241 |
+
-respath ./workdir/dragon \
|
242 |
+
-d 8023
|
243 |
+
```
|
244 |
+
|
245 |
+
### Case: Unicorn
|
246 |
+
|
247 |
+
**Preview:**
|
248 |
+
|
249 |
+
| <img src="./img/unicorn/svg_iter0.svg"> | <img src="./img/unicorn/svg_iter100.svg"> | <img src="./img/unicorn/visual_best.svg"> |
|
250 |
+
|-----------------------------------------|-------------------------------------------|-------------------------------------------|
|
251 |
+
| Strokes Initialization | 100 step | 1200 step |
|
252 |
+
|
253 |
+
**Script:**
|
254 |
+
|
255 |
+
```shell
|
256 |
+
python run_painterly_render.py \
|
257 |
+
-c diffsketcher.yaml \
|
258 |
+
-eval_step 10 -save_step 10 \
|
259 |
+
-update "token_ind=2" \
|
260 |
+
-pt "A unicorn is running on the grassland" \
|
261 |
+
-respath ./workdir/unicorn \
|
262 |
+
-d 914678
|
263 |
+
```
|
264 |
+
|
265 |
+
**Preview:**
|
266 |
+
|
267 |
+
| <img src="./img/unicorn-2/svg_iter0.svg"> | <img src="./img/unicorn-2/svg_iter100.svg"> | <img src="./img/unicorn-2/visual_best.svg"> |
|
268 |
+
|-------------------------------------------|---------------------------------------------|---------------------------------------------|
|
269 |
+
| Strokes Initialization | 100 step | 1240 step |
|
270 |
+
|
271 |
+
**Script:**
|
272 |
+
|
273 |
+
```shell
|
274 |
+
python run_painterly_render.py \
|
275 |
+
-c diffsketcher.yaml \
|
276 |
+
-eval_step 10 -save_step 10 \
|
277 |
+
-update "token_ind=2 softmax_temp=0.5 mask_object=True" \
|
278 |
+
-pt "A unicorn is running on the grassland" \
|
279 |
+
-respath ./workdir/unicorn \
|
280 |
+
-d 9998
|
281 |
+
```
|
282 |
+
|
283 |
+
### Case: Mushroom
|
284 |
+
|
285 |
+
**Preview:**
|
286 |
+
|
287 |
+
| <img src="./img/Mushroom/svg_iter0.svg"> | <img src="./img/Mushroom/svg_iter100.svg"> | <img src="./img/Mushroom/final_svg.svg"> |
|
288 |
+
|------------------------------------------|--------------------------------------------|------------------------------------------|
|
289 |
+
| Strokes Initialization | 100 step | 2000 step |
|
290 |
+
|
291 |
+
**Script:**
|
292 |
+
|
293 |
+
```shell
|
294 |
+
CUDA_VISIBLE_DEVICES=0 python run_painterly_render.py \
|
295 |
+
-c diffsketcher.yaml \
|
296 |
+
-eval_step 10 -save_step 10 \
|
297 |
+
-update "token_ind=4 num_paths=84 comp_idx=0 attn_coeff=1 softmax_temp=0.4 xdog_intersec=False sds.num_aug=4 sds.grad_scale=2 sds.warmup=0 clip.vis_loss=0 clip.num_aug=4 clip.text_visual_coeff=0 perceptual.coeff=0 opacity_delta=0.2 lr_scheduler=True num_iter=2000" \
|
298 |
+
-pt "a brightly colored mushroom growing on a logt, minimal 2d line drawing. trending on artstation." \
|
299 |
+
-respath ./workdir/Mushroom \
|
300 |
+
-d 621024
|
301 |
+
```
|