# -------------------------------------------------------- # PersonalizeSAM -- Personalize Segment Anything Model with One Shot # Licensed under The MIT License [see LICENSE for details] # -------------------------------------------------------- from PIL import Image import torch import torch.nn as nn import gradio as gr import numpy as np from torch.nn import functional as F from show import * from per_segment_anything import sam_model_registry, SamPredictor class Mask_Weights(nn.Module): def __init__(self): super().__init__() self.weights = nn.Parameter(torch.ones(2, 1, requires_grad=True) / 3) def point_selection(mask_sim, topk=1): # Top-1 point selection w, h = mask_sim.shape topk_xy = mask_sim.flatten(0).topk(topk)[1] topk_x = (topk_xy // h).unsqueeze(0) topk_y = (topk_xy - topk_x * h) topk_xy = torch.cat((topk_y, topk_x), dim=0).permute(1, 0) topk_label = np.array([1] * topk) topk_xy = topk_xy.cpu().numpy() # Top-last point selection last_xy = mask_sim.flatten(0).topk(topk, largest=False)[1] last_x = (last_xy // h).unsqueeze(0) last_y = (last_xy - last_x * h) last_xy = torch.cat((last_y, last_x), dim=0).permute(1, 0) last_label = np.array([0] * topk) last_xy = last_xy.cpu().numpy() return topk_xy, topk_label, last_xy, last_label def calculate_dice_loss(inputs, targets, num_masks = 1): """ Compute the DICE loss, similar to generalized IOU for masks Args: inputs: A float tensor of arbitrary shape. The predictions for each example. targets: A float tensor with the same shape as inputs. Stores the binary classification label for each element in inputs (0 for the negative class and 1 for the positive class). """ inputs = inputs.sigmoid() inputs = inputs.flatten(1) numerator = 2 * (inputs * targets).sum(-1) denominator = inputs.sum(-1) + targets.sum(-1) loss = 1 - (numerator + 1) / (denominator + 1) return loss.sum() / num_masks def calculate_sigmoid_focal_loss(inputs, targets, num_masks = 1, alpha: float = 0.25, gamma: float = 2): """ Loss used in RetinaNet for dense detection: https://arxiv.org/abs/1708.02002. Args: inputs: A float tensor of arbitrary shape. The predictions for each example. targets: A float tensor with the same shape as inputs. Stores the binary classification label for each element in inputs (0 for the negative class and 1 for the positive class). alpha: (optional) Weighting factor in range (0,1) to balance positive vs negative examples. Default = -1 (no weighting). gamma: Exponent of the modulating factor (1 - p_t) to balance easy vs hard examples. Returns: Loss tensor """ prob = inputs.sigmoid() ce_loss = F.binary_cross_entropy_with_logits(inputs, targets, reduction="none") p_t = prob * targets + (1 - prob) * (1 - targets) loss = ce_loss * ((1 - p_t) ** gamma) if alpha >= 0: alpha_t = alpha * targets + (1 - alpha) * (1 - targets) loss = alpha_t * loss return loss.mean(1).sum() / num_masks def inference(ic_image, ic_mask, image1, image2): # in context image and mask ic_image = np.array(ic_image.convert("RGB")) ic_mask = np.array(ic_mask.convert("RGB")) sam_type, sam_ckpt = 'vit_h', 'sam_vit_h_4b8939.pth' sam = sam_model_registry[sam_type](checkpoint=sam_ckpt).cuda() # sam = sam_model_registry[sam_type](checkpoint=sam_ckpt) predictor = SamPredictor(sam) # Image features encoding ref_mask = predictor.set_image(ic_image, ic_mask) ref_feat = predictor.features.squeeze().permute(1, 2, 0) ref_mask = F.interpolate(ref_mask, size=ref_feat.shape[0: 2], mode="bilinear") ref_mask = ref_mask.squeeze()[0] # Target feature extraction print("======> Obtain Location Prior" ) target_feat = ref_feat[ref_mask > 0] target_embedding = target_feat.mean(0).unsqueeze(0) target_feat = target_embedding / target_embedding.norm(dim=-1, keepdim=True) target_embedding = target_embedding.unsqueeze(0) output_image = [] for test_image in [image1, image2]: print("======> Testing Image" ) test_image = np.array(test_image.convert("RGB")) # Image feature encoding predictor.set_image(test_image) test_feat = predictor.features.squeeze() # Cosine similarity C, h, w = test_feat.shape test_feat = test_feat / test_feat.norm(dim=0, keepdim=True) test_feat = test_feat.reshape(C, h * w) sim = target_feat @ test_feat sim = sim.reshape(1, 1, h, w) sim = F.interpolate(sim, scale_factor=4, mode="bilinear") sim = predictor.model.postprocess_masks( sim, input_size=predictor.input_size, original_size=predictor.original_size).squeeze() # Positive-negative location prior topk_xy_i, topk_label_i, last_xy_i, last_label_i = point_selection(sim, topk=1) topk_xy = np.concatenate([topk_xy_i, last_xy_i], axis=0) topk_label = np.concatenate([topk_label_i, last_label_i], axis=0) # Obtain the target guidance for cross-attention layers sim = (sim - sim.mean()) / torch.std(sim) sim = F.interpolate(sim.unsqueeze(0).unsqueeze(0), size=(64, 64), mode="bilinear") attn_sim = sim.sigmoid_().unsqueeze(0).flatten(3) # First-step prediction masks, scores, logits, _ = predictor.predict( point_coords=topk_xy, point_labels=topk_label, multimask_output=False, attn_sim=attn_sim, # Target-guided Attention target_embedding=target_embedding # Target-semantic Prompting ) best_idx = 0 # Cascaded Post-refinement-1 masks, scores, logits, _ = predictor.predict( point_coords=topk_xy, point_labels=topk_label, mask_input=logits[best_idx: best_idx + 1, :, :], multimask_output=True) best_idx = np.argmax(scores) # Cascaded Post-refinement-2 y, x = np.nonzero(masks[best_idx]) x_min = x.min() x_max = x.max() y_min = y.min() y_max = y.max() input_box = np.array([x_min, y_min, x_max, y_max]) masks, scores, logits, _ = predictor.predict( point_coords=topk_xy, point_labels=topk_label, box=input_box[None, :], mask_input=logits[best_idx: best_idx + 1, :, :], multimask_output=True) best_idx = np.argmax(scores) final_mask = masks[best_idx] mask_colors = np.zeros((final_mask.shape[0], final_mask.shape[1], 3), dtype=np.uint8) mask_colors[final_mask, :] = np.array([[128, 0, 0]]) output_image.append(Image.fromarray((mask_colors * 0.6 + test_image * 0.4).astype('uint8'), 'RGB')) return output_image[0].resize((224, 224)), output_image[1].resize((224, 224)) def inference_scribble(image, image1, image2): # in context image and mask ic_image = image["image"] ic_mask = image["mask"] ic_image = np.array(ic_image.convert("RGB")) ic_mask = np.array(ic_mask.convert("RGB")) sam_type, sam_ckpt = 'vit_h', 'sam_vit_h_4b8939.pth' sam = sam_model_registry[sam_type](checkpoint=sam_ckpt).cuda() # sam = sam_model_registry[sam_type](checkpoint=sam_ckpt) predictor = SamPredictor(sam) # Image features encoding ref_mask = predictor.set_image(ic_image, ic_mask) ref_feat = predictor.features.squeeze().permute(1, 2, 0) ref_mask = F.interpolate(ref_mask, size=ref_feat.shape[0: 2], mode="bilinear") ref_mask = ref_mask.squeeze()[0] # Target feature extraction print("======> Obtain Location Prior" ) target_feat = ref_feat[ref_mask > 0] target_embedding = target_feat.mean(0).unsqueeze(0) target_feat = target_embedding / target_embedding.norm(dim=-1, keepdim=True) target_embedding = target_embedding.unsqueeze(0) output_image = [] for test_image in [image1, image2]: print("======> Testing Image" ) test_image = np.array(test_image.convert("RGB")) # Image feature encoding predictor.set_image(test_image) test_feat = predictor.features.squeeze() # Cosine similarity C, h, w = test_feat.shape test_feat = test_feat / test_feat.norm(dim=0, keepdim=True) test_feat = test_feat.reshape(C, h * w) sim = target_feat @ test_feat sim = sim.reshape(1, 1, h, w) sim = F.interpolate(sim, scale_factor=4, mode="bilinear") sim = predictor.model.postprocess_masks( sim, input_size=predictor.input_size, original_size=predictor.original_size).squeeze() # Positive-negative location prior topk_xy_i, topk_label_i, last_xy_i, last_label_i = point_selection(sim, topk=1) topk_xy = np.concatenate([topk_xy_i, last_xy_i], axis=0) topk_label = np.concatenate([topk_label_i, last_label_i], axis=0) # Obtain the target guidance for cross-attention layers sim = (sim - sim.mean()) / torch.std(sim) sim = F.interpolate(sim.unsqueeze(0).unsqueeze(0), size=(64, 64), mode="bilinear") attn_sim = sim.sigmoid_().unsqueeze(0).flatten(3) # First-step prediction masks, scores, logits, _ = predictor.predict( point_coords=topk_xy, point_labels=topk_label, multimask_output=False, attn_sim=attn_sim, # Target-guided Attention target_embedding=target_embedding # Target-semantic Prompting ) best_idx = 0 # Cascaded Post-refinement-1 masks, scores, logits, _ = predictor.predict( point_coords=topk_xy, point_labels=topk_label, mask_input=logits[best_idx: best_idx + 1, :, :], multimask_output=True) best_idx = np.argmax(scores) # Cascaded Post-refinement-2 y, x = np.nonzero(masks[best_idx]) x_min = x.min() x_max = x.max() y_min = y.min() y_max = y.max() input_box = np.array([x_min, y_min, x_max, y_max]) masks, scores, logits, _ = predictor.predict( point_coords=topk_xy, point_labels=topk_label, box=input_box[None, :], mask_input=logits[best_idx: best_idx + 1, :, :], multimask_output=True) best_idx = np.argmax(scores) final_mask = masks[best_idx] mask_colors = np.zeros((final_mask.shape[0], final_mask.shape[1], 3), dtype=np.uint8) mask_colors[final_mask, :] = np.array([[128, 0, 0]]) output_image.append(Image.fromarray((mask_colors * 0.6 + test_image * 0.4).astype('uint8'), 'RGB')) return output_image[0].resize((224, 224)), output_image[1].resize((224, 224)) def inference_finetune(ic_image, ic_mask, image1, image2): # in context image and mask ic_image = np.array(ic_image.convert("RGB")) ic_mask = np.array(ic_mask.convert("RGB")) gt_mask = torch.tensor(ic_mask)[:, :, 0] > 0 gt_mask = gt_mask.float().unsqueeze(0).flatten(1).cuda() # gt_mask = gt_mask.float().unsqueeze(0).flatten(1) sam_type, sam_ckpt = 'vit_h', 'sam_vit_h_4b8939.pth' sam = sam_model_registry[sam_type](checkpoint=sam_ckpt).cuda() # sam = sam_model_registry[sam_type](checkpoint=sam_ckpt) for name, param in sam.named_parameters(): param.requires_grad = False predictor = SamPredictor(sam) print("======> Obtain Self Location Prior" ) # Image features encoding ref_mask = predictor.set_image(ic_image, ic_mask) ref_feat = predictor.features.squeeze().permute(1, 2, 0) ref_mask = F.interpolate(ref_mask, size=ref_feat.shape[0: 2], mode="bilinear") ref_mask = ref_mask.squeeze()[0] # Target feature extraction target_feat = ref_feat[ref_mask > 0] target_feat_mean = target_feat.mean(0) target_feat_max = torch.max(target_feat, dim=0)[0] target_feat = (target_feat_max / 2 + target_feat_mean / 2).unsqueeze(0) # Cosine similarity h, w, C = ref_feat.shape target_feat = target_feat / target_feat.norm(dim=-1, keepdim=True) ref_feat = ref_feat / ref_feat.norm(dim=-1, keepdim=True) ref_feat = ref_feat.permute(2, 0, 1).reshape(C, h * w) sim = target_feat @ ref_feat sim = sim.reshape(1, 1, h, w) sim = F.interpolate(sim, scale_factor=4, mode="bilinear") sim = predictor.model.postprocess_masks( sim, input_size=predictor.input_size, original_size=predictor.original_size).squeeze() # Positive location prior topk_xy, topk_label, _, _ = point_selection(sim, topk=1) print('======> Start Training') # Learnable mask weights mask_weights = Mask_Weights().cuda() # mask_weights = Mask_Weights() mask_weights.train() train_epoch = 100 optimizer = torch.optim.AdamW(mask_weights.parameters(), lr=1e-3, eps=1e-4) scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, train_epoch) for train_idx in range(train_epoch): # Run the decoder masks, scores, logits, logits_high = predictor.predict( point_coords=topk_xy, point_labels=topk_label, multimask_output=True) logits_high = logits_high.flatten(1) # Weighted sum three-scale masks weights = torch.cat((1 - mask_weights.weights.sum(0).unsqueeze(0), mask_weights.weights), dim=0) logits_high = logits_high * weights logits_high = logits_high.sum(0).unsqueeze(0) dice_loss = calculate_dice_loss(logits_high, gt_mask) focal_loss = calculate_sigmoid_focal_loss(logits_high, gt_mask) loss = dice_loss + focal_loss optimizer.zero_grad() loss.backward() optimizer.step() scheduler.step() if train_idx % 10 == 0: print('Train Epoch: {:} / {:}'.format(train_idx, train_epoch)) current_lr = scheduler.get_last_lr()[0] print('LR: {:.6f}, Dice_Loss: {:.4f}, Focal_Loss: {:.4f}'.format(current_lr, dice_loss.item(), focal_loss.item())) mask_weights.eval() weights = torch.cat((1 - mask_weights.weights.sum(0).unsqueeze(0), mask_weights.weights), dim=0) weights_np = weights.detach().cpu().numpy() print('======> Mask weights:\n', weights_np) print('======> Start Testing') output_image = [] for test_image in [image1, image2]: test_image = np.array(test_image.convert("RGB")) # Image feature encoding predictor.set_image(test_image) test_feat = predictor.features.squeeze() # Image feature encoding predictor.set_image(test_image) test_feat = predictor.features.squeeze() # Cosine similarity C, h, w = test_feat.shape test_feat = test_feat / test_feat.norm(dim=0, keepdim=True) test_feat = test_feat.reshape(C, h * w) sim = target_feat @ test_feat sim = sim.reshape(1, 1, h, w) sim = F.interpolate(sim, scale_factor=4, mode="bilinear") sim = predictor.model.postprocess_masks( sim, input_size=predictor.input_size, original_size=predictor.original_size).squeeze() # Positive location prior topk_xy, topk_label, _, _ = point_selection(sim, topk=1) # First-step prediction masks, scores, logits, logits_high = predictor.predict( point_coords=topk_xy, point_labels=topk_label, multimask_output=True) # Weighted sum three-scale masks logits_high = logits_high * weights.unsqueeze(-1) logit_high = logits_high.sum(0) mask = (logit_high > 0).detach().cpu().numpy() logits = logits * weights_np[..., None] logit = logits.sum(0) # Cascaded Post-refinement-1 y, x = np.nonzero(mask) x_min = x.min() x_max = x.max() y_min = y.min() y_max = y.max() input_box = np.array([x_min, y_min, x_max, y_max]) masks, scores, logits, _ = predictor.predict( point_coords=topk_xy, point_labels=topk_label, box=input_box[None, :], mask_input=logit[None, :, :], multimask_output=True) best_idx = np.argmax(scores) # Cascaded Post-refinement-2 y, x = np.nonzero(masks[best_idx]) x_min = x.min() x_max = x.max() y_min = y.min() y_max = y.max() input_box = np.array([x_min, y_min, x_max, y_max]) masks, scores, logits, _ = predictor.predict( point_coords=topk_xy, point_labels=topk_label, box=input_box[None, :], mask_input=logits[best_idx: best_idx + 1, :, :], multimask_output=True) best_idx = np.argmax(scores) final_mask = masks[best_idx] mask_colors = np.zeros((final_mask.shape[0], final_mask.shape[1], 3), dtype=np.uint8) mask_colors[final_mask, :] = np.array([[128, 0, 0]]) output_image.append(Image.fromarray((mask_colors * 0.6 + test_image * 0.4).astype('uint8'), 'RGB')) return output_image[0].resize((224, 224)), output_image[1].resize((224, 224)) description = """
[Github] [Paper]
""" main = gr.Interface( fn=inference, inputs=[ gr.Image(label="one-shot image", type='pil'), gr.Image(label="one-shot mask", type='pil'), gr.Image(label="test image1", type='pil'), gr.Image(label="test image2", type='pil'), ], outputs=[ gr.Image(label="output image1", type='pil').style(height=256, width=256), gr.Image(label="output image2", type='pil').style(height=256, width=256), ], allow_flagging="never", cache_examples=False, title="Personalize Segment Anything Model with 1 Shot", description=description, examples=[ ["./examples/cat_00.jpg", "./examples/cat_00.png", "./examples/cat_01.jpg", "./examples/cat_02.jpg"], ["./examples/colorful_sneaker_00.jpg", "./examples/colorful_sneaker_00.png", "./examples/colorful_sneaker_01.jpg", "./examples/colorful_sneaker_02.jpg"], ["./examples/duck_toy_00.jpg", "./examples/duck_toy_00.png", "./examples/duck_toy_01.jpg", "./examples/duck_toy_02.jpg"], ] ) main_scribble = gr.Interface( fn=inference_scribble, inputs=[ gr.ImageMask(label="[Stroke] Draw on Image", brush_radius=4, type='pil'), gr.Image(label="test image1", type='pil'), gr.Image(label="test image2", type='pil'), ], outputs=[ gr.Image(label="output image1", type='pil').style(height=256, width=256), gr.Image(label="output image2", type='pil').style(height=256, width=256), ], allow_flagging="never", cache_examples=False, title="Personalize Segment Anything Model with 1 Shot", description=description, examples=[ ["./examples/cat_00.jpg", "./examples/cat_01.jpg", "./examples/cat_02.jpg"], ["./examples/colorful_sneaker_00.jpg", "./examples/colorful_sneaker_01.jpg", "./examples/colorful_sneaker_02.jpg"], ["./examples/duck_toy_00.jpg", "./examples/duck_toy_01.jpg", "./examples/duck_toy_02.jpg"], ] ) main_finetune = gr.Interface( fn=inference_finetune, inputs=[ gr.Image(label="one-shot image", type='pil'), gr.Image(label="one-shot mask", type='pil'), gr.Image(label="test image1", type='pil'), gr.Image(label="test image2", type='pil'), ], outputs=[ gr.Image(label="output image1", type='pil').style(height=256, width=256), gr.Image(label="output image2", type='pil').style(height=256, width=256), ], allow_flagging="never", cache_examples=False, title="Personalize Segment Anything Model with 1 Shot", description=description, examples=[ ["./examples/cat_00.jpg", "./examples/cat_00.png", "./examples/cat_01.jpg", "./examples/cat_02.jpg"], ["./examples/colorful_sneaker_00.jpg", "./examples/colorful_sneaker_00.png", "./examples/colorful_sneaker_01.jpg", "./examples/colorful_sneaker_02.jpg"], ["./examples/duck_toy_00.jpg", "./examples/duck_toy_00.png", "./examples/duck_toy_01.jpg", "./examples/duck_toy_02.jpg"], ] ) demo = gr.Blocks() with demo: gr.TabbedInterface( [main, main_scribble, main_finetune], ["PerSAM", " PerSAM-Scribble", "PerSAM-F"], ) demo.launch(enable_queue=False)