mistake0316/style-transfer-clip 🔢📝🖼️✓ → ❓

▶️ 1.3K runs 📅 Nov 2021 ⚙️ Cog 0.1.3+shimmed 🔗 GitHub
clip-loss image-stylization style-transfer text-to-image

About

Guide Style Transfer with CLIP loss

Example Output

Output

[object Object][object Object][object Object][object Object][object Object][object Object][object Object][object Object][object Object][object Object]

Performance Metrics

66.12s Total Time
All Input Parameters
{
  "lr": "0.01",
  "text": "pancake",
  "image": "https://replicate.delivery/mgxm/0f51aae9-ef91-4ee6-a83f-065abbb7a65c/doge.jpeg",
  "resize_flag": true,
  "optimize_steps": "100",
  "output_aug_flag": true,
  "output_aug_time": 32,
  "center_crop_flag": true,
  "display_every_step": "10",
  "short_edge_target_len": "384"
}
Input Parameters
lr Type: numberDefault: 0.01
text Type: string
prompt text
image Type: string
input image
resize_flag Type: booleanDefault: true
resize image or not
optimize_steps Type: integerDefault: 100
total steps
output_aug_flag Type: booleanDefault: true
augment before compute clip loss
output_aug_time Type: integerDefault: 8
augment times if output_aug_flag is True
center_crop_flag Type: booleanDefault: true
center crop content image or not
display_every_step Type: integerDefault: 10
short_edge_target_len Type: integerDefault: 384
the short edge size if resize_flag is True
Output Schema

Type: arrayItems Type: object

Example Execution Logs
loading content
pancake
tensor([[49406, 18723, 49407,     0,     0,     0,     0,     0,     0,     0,
             0,     0,     0,     0,     0,     0,     0,     0,     0,     0,
             0,     0,     0,     0,     0,     0,     0,     0,     0,     0,
             0,     0,     0,     0,     0,     0,     0,     0,     0,     0,
             0,     0,     0,     0,     0,     0,     0,     0,     0,     0,
             0,     0,     0,     0,     0,     0,     0,     0,     0,     0,
             0,     0,     0,     0,     0,     0,     0,     0,     0,     0,
             0,     0,     0,     0,     0,     0,     0]], device='cuda:0')
step : 0
	loss :  0.7568359375
step : 10
	loss :  0.6923828125
step : 20
	loss :  0.66748046875
step : 30
	loss :  0.6591796875
step : 40
	loss :  0.65283203125
step : 50
	loss :  0.6474609375
step : 60
	loss :  0.6455078125
step : 70
	loss :  0.64208984375
step : 80
	loss :  0.640625
step : 90
	loss :  0.6376953125
step : 99
	loss :  0.63916015625
Version Details
Version ID
4f26fddbeddab99fdbd1f4a38884f7f1fa70e015a89071583af82aec8d39b50a
Version Created
November 25, 2021
Run on Replicate →