-
Notifications
You must be signed in to change notification settings - Fork 72
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
1 parent
0e03fc5
commit 82fbeba
Showing
18 changed files
with
781 additions
and
88 deletions.
There are no files selected for viewing
Binary file not shown.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Empty file.
169 changes: 169 additions & 0 deletions
169
app/src/algorithm/video_optimize_local_processing/video_optimize_local_processing.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,169 @@ | ||
| import os | ||
| import shutil | ||
|
|
||
| import cv2 | ||
| import numpy as np | ||
| from PIL import Image,ImageFont,ImageDraw | ||
| from app.src.utils.logger import Logger | ||
| from moviepy.editor import VideoFileClip, ImageSequenceClip | ||
|
|
||
| from app.src.algorithm.base.sam2.video_predictor import VideoPredictor | ||
|
|
||
|
|
||
| class LocalModel: | ||
| def __init__(self, config, logger): | ||
| self.logger = logger | ||
| self.config = config | ||
| self.size = config["input"]["size"] | ||
| self.lama_model = None | ||
| self.sam2_video_model = None | ||
| self.sam2_image_model = None | ||
| self.point_diameter = int(min(self.size) * 0.05) | ||
|
|
||
| def initial(self): | ||
| self.logger.write_log("interval:1:1:1:0:Video Initial") | ||
| if self.sam2_video_model is None: | ||
| self.sam2_video_model = VideoPredictor(self.config["config"]["sam2"], self.logger) | ||
| self.sam2_video_model.set_video(self.config["input"]["video_frame_path"]) | ||
| self.logger.write_log("interval:1:1:1:0:Video Initial") | ||
|
|
||
| def reset(self): | ||
| self.sam2_video_model.reset() | ||
|
|
||
| def add_point(self, point, label, ann_frame_idx): | ||
| if self.sam2_video_model is None: | ||
| self.sam2_video_model = VideoPredictor(self.config["config"]["sam2"], self.logger) | ||
| self.sam2_video_model.set_video(self.config["input"]["video_frame_path"]) | ||
| out_obj_ids, out_mask_logits = self.sam2_video_model.add_point( | ||
| point, | ||
| label, | ||
| ann_frame_idx | ||
| ) | ||
| return out_obj_ids, out_mask_logits | ||
|
|
||
| def show_segment_frame(self, mask_logits, ann_frame_idx): | ||
| frame_path = os.path.join(self.config["input"]["video_frame_path"], self.sam2_video_model.frame_names[ann_frame_idx]) | ||
| frame = Image.open(frame_path) | ||
| mask_color = self.config["input"]["mask_color"] | ||
| mask_color = Image.new("RGBA", (self.size[0], self.size[1]), mask_color) | ||
| mask_color = np.array(mask_color) | ||
| mask_logits = (mask_logits[0] > 0.0).cpu().numpy() | ||
| mask_logits = mask_logits.reshape((self.size[1], self.size[0], -1)) | ||
| mask_rgba = Image.fromarray(mask_color * mask_logits) | ||
|
|
||
| show_frame = Image.alpha_composite(frame.convert('RGBA'), mask_rgba).convert("RGB") | ||
| return show_frame | ||
|
|
||
| def _get_mask(self, mask_image, mask_logit, size): | ||
| w, h = size | ||
| mask_logit = mask_logit.reshape(h, w, 1) * 255 | ||
| mask_image = np.logical_or(mask_image, mask_logit) | ||
| return mask_image | ||
|
|
||
| def processing(self, process_type, output_path): | ||
| self.logger.write_log("interval:3:1:1:0") | ||
| if self.sam2_video_model is None: | ||
| self.sam2_video_model = VideoPredictor(self.config["config"]["sam2"], self.logger) | ||
| self.sam2_video_model.set_video(self.config["input"]["video_frame_path"]) | ||
| video_segments = self.sam2_video_model.propagate_video() | ||
| self.logger.write_log("interval:3:1:1:1") | ||
|
|
||
| process_temp_dir = self.config["output"]["process_temp_dir"] | ||
| image_paths = list() | ||
| process_num = len(video_segments) | ||
| self.logger.write_log(f"follow:3:2:{process_num}:0") | ||
| for out_frame_idx, item in video_segments.items(): | ||
| mask_image = np.zeros((self.size[1], self.size[0], 1)) | ||
| for out_obj_id, out_mask in item.items(): | ||
| mask_image = self._get_mask(mask_image, out_mask, self.size) | ||
|
|
||
| frame_path = os.path.join(self.config["input"]["video_frame_path"], self.sam2_video_model.frame_names[out_frame_idx]) | ||
| name, _ = self.sam2_video_model.frame_names[out_frame_idx].rsplit(".", 1) | ||
| temp_image_path = os.path.join(process_temp_dir, f"{name}.png") | ||
|
|
||
| frame = Image.open(frame_path) | ||
| gray_frame = frame.convert("L").convert("RGB") | ||
| color_frame = frame | ||
| if process_type == "gray": | ||
| background = np.array(color_frame) | ||
| foreground = np.array(gray_frame) | ||
| else: | ||
| background = np.array(gray_frame) | ||
| foreground = np.array(color_frame) | ||
|
|
||
| if len(item.items()) > 0: | ||
| output_image = mask_image * foreground + (1 - mask_image) * background | ||
| Image.fromarray(np.uint8(output_image)).save(temp_image_path) | ||
| else: | ||
| Image.fromarray(background).save(temp_image_path) | ||
| self.logger.write_log(f"follow:3:2:{process_num}:{out_frame_idx+1}") | ||
| image_paths.append(temp_image_path) | ||
| self.logger.write_log(f"follow:3:2:{process_num}:{process_num}") | ||
|
|
||
| self.logger.write_log(f"interval:3:3:1:0") | ||
| video_path = self.config["input"]["video_path"] | ||
| video = VideoFileClip(video_path) | ||
| fps = video.fps | ||
| output_video = ImageSequenceClip(image_paths, fps=fps) | ||
| output_video = output_video.set_audio(video.audio) | ||
| output_video.write_videofile(output_path) | ||
| shutil.rmtree(process_temp_dir, ignore_errors=True) | ||
| video.close() | ||
| self.logger.write_log(f"interval:3:3:1:1") | ||
|
|
||
|
|
||
| def video_optimize_local_processing(input_data, local_model=None): | ||
|
|
||
| timestamp = input_data["input"]["timestamp"] | ||
| log_path = input_data["input"]["log_path"] | ||
| logger = Logger(log_path, timestamp) | ||
| if local_model is None: | ||
| local_model = LocalModel(input_data, logger) | ||
| else: | ||
| config = local_model.config | ||
| config.update(input_data) | ||
| local_model.config = config | ||
|
|
||
| opt_type = input_data["type"] | ||
| if opt_type == "add": | ||
| # Add a point | ||
| prompt = input_data["input"]["prompt"] | ||
| _, out_mask_logits = local_model.add_point( | ||
| prompt["data"], | ||
| prompt["value"], | ||
| input_data["input"]["ann_frame_idx"], | ||
| ) | ||
| show_frame = local_model.show_segment_frame(out_mask_logits, input_data["input"]["ann_frame_idx"]) | ||
| show_frame.save(input_data["output"]["show_temp_image"]) | ||
| elif opt_type == "remove": | ||
| # Remove a point | ||
| prompts = input_data["input"]["prompts"] | ||
| ann_frame_idx = input_data["input"]["ann_frame_idx"] | ||
| local_model.reset() | ||
| out_mask_logits = None | ||
| for frame_id, frame_prompts in prompts.items(): | ||
| if frame_id == ann_frame_idx: | ||
| for prompt in frame_prompts: | ||
| if prompt["type"] == "point": | ||
| _, out_mask_logits = local_model.add_point( | ||
| np.array(prompt["data"]), | ||
| np.array([prompt["value"]]), | ||
| frame_id, | ||
| ) | ||
| else: | ||
| for prompt in frame_prompts: | ||
| if prompt["type"] == "point": | ||
| _, _ = local_model.add_point( | ||
| np.array(prompt["data"]), | ||
| np.array([prompt["value"]]), | ||
| frame_id, | ||
| ) | ||
| if out_mask_logits is not None: | ||
| show_frame = local_model.show_segment_frame(out_mask_logits, input_data["input"]["ann_frame_idx"]) | ||
| show_frame.save(input_data["output"]["show_temp_image"]) | ||
| elif opt_type == "initial": | ||
| local_model.initial() | ||
| else: | ||
| local_model.processing(input_data["input"]["process_type"], input_data["output"]["video_path"]) | ||
| return local_model | ||
|
|
Oops, something went wrong.