Video_sampling scripts added by yashsuman15 · Pull Request #16 · Labellerr/SDKPython

yashsuman15 · 2025-10-02T11:47:59Z

created SDKPython subdir in labellerr
added utils/client_utils
added video_sampling/pyscene_detect.py

test: Add integration tests for Labellerr SDK This commit introduces integration tests for core SDK functionalities: Features tested: - Project creation with multiple annotation types: * Polygon annotations * Bounding box detection * Classification (select, dropdown, radio) * Text input fields * Combined annotation types - Project export functionality - Pre-annotation upload support (COCO JSON format) Test structure: /tests /integration - Create_Project.py # Project creation test cases - Export_project.py # Export functionality tests - Pre_annotation.py # Pre-annotation upload tests - main.py # Test runner - cred.py # Credentials config (gitignored) Requirements: - Valid API credentials in cred.py - Test image dataset in test_img/ - Sample annotations in annotations.json Note: Remember to update cred.py with valid credentials before running tests

-removed __pychace__ -added .gitignore

- added utils/client_utils - added video_sampling/pyscene_detect.py

- add videointelligence module in requirements

ximihoque · 2025-10-04T09:43:35Z

+import os
+import cv2
+from PIL import Image
+from dataclasses import dataclass, asdict


use pydantic instead

ximihoque · 2025-10-04T09:49:56Z

+        mapping = {
+            "file_id": result.file_id,
+            "output_folder": result.output_folder,
+            "total_frames": result.total_frames,
+            "total_selected_frames": len(result.selected_frames),
+            "detection_method": "Google Cloud Video Intelligence API - Shot Change Detection",


use pydantic's model dump

ximihoque · 2025-10-04T09:50:43Z

+    selected_frames: List[SceneFrame]
+
+
+class GeminiSceneDetect:


create all singleton classes

ximihoque · 2025-10-04T09:55:05Z

+#         video_path=video_path,
+#         file_id="video_001",
+#         credentials_path=cred_json_path  # Uses GOOGLE_APPLICATION_CREDENTIALS env var
+#     )


detector should know the attrs of the labellerr_file, so that if it requires, video_path, file_id, or frames etc whatever, it can access from the labellerr_file instance itself

ximihoque · 2025-10-04T09:56:32Z

+class GeminiSceneDetect:
+    """Google Cloud Video Intelligence API scene detection and frame extraction."""
+
+    def __init__(self, video_path: str, file_id: str, gcs_uri: Optional[str] = None, credentials_path: Optional[str] = None):


gcs_uri shouldn't be required in the constructor

ximihoque · 2025-10-04T10:03:30Z

+        """
+        features = [videointelligence.Feature.SHOT_CHANGE_DETECTION]
+
+        if self.gcs_uri:


gcs uri will only be used if the user is using Labellerr's gemini model. For files > 10 mb, and if user is using his own svc account, then we raise an error.

ximihoque · 2025-10-04T10:06:46Z

+            credentials_path: Path to service account JSON key file.
+                            If None, uses GOOGLE_APPLICATION_CREDENTIALS environment variable


if any svc account file is provided then 10 mb limit is enforced, if not provided then copy the video file to gcs, then send the gcs_uri path to gemini (to be done, when we move this code to the DataOps repo)

ximihoque · 2025-10-04T10:13:55Z

+            success_count = 0
+            failed_frames = []
+
+            print(f"Downloading {len(frames_data)} frames to: {save_path}")


use multithreading to download all the frames

ximihoque · 2025-10-04T10:16:42Z

+# https://api.labellerr.com/data/file_data?file_id=c44f38f6-0186-436f-8c2d-ffb50a539c76&include_answers=false&project_id=gabrila_artificial_duck_74237&uuid=1d4c9b58-c6a4-4ca8-9583-b6b6cd25ef12
+
+
+class FileMetadataService:


move this to services/labellerr_files/init.py

ximihoque · 2025-10-04T10:20:53Z

+import subprocess
+import os
+
+class FFMPEG:


make it FFMPEGSceneDetect

…SDKPython into feature/LABIMP-7672

yashsuman15

Modify the scripts based on comments

This reverts commit 5addbaa.

ximihoque · 2025-10-07T04:44:00Z

add versioninig to each requirement

yashsuman15

can you help me how can i implement limit in scene detections file based on density (which you have me told earlier)

currently all the scene change are extracted and i need to select only specific numbers of that.

added the updated client.py added the restructure files

Refactor video sampling services to include dataset ID in output folder structure - Updated FFMPEGSceneDetect to derive dataset ID from video path and include it in the output folder. - Modified PySceneDetect to incorporate dataset ID in the output folder structure. - Adjusted SSIMSceneDetect to also use dataset ID for organizing output folders.

ximihoque · 2025-10-13T02:46:02Z

Comments on the notebook -

It should be dataset = LabellerrDataset(client, dataset_id, project_id)
results = dataset.download()

Where is the boilerplate code for using the frames to create ImageDataset from the frames ?

Write docstrings and pseudo code to create ImageDataset with all the sampled frames.
Use that dataset to create a Project. Ensure the file names are file_id/frame_idx in the dataset for backtracking.
Create an export of this project
Using the export, upload pre-annotations to the VideoProject [ this feature will be available today ].

…ataset and update related methods in SDK notebook - Updated import statements in __init__.py and SDK.ipynb to reflect the new class name. - Renamed method process_all_videos() to download() in LabellerrDataset class. - Adjusted SDK notebook to use the new download() method for processing videos. - Corrected dataset directory path in SDK notebook. - Added sections for project creation and image dataset handling in the SDK notebook.

yashsuman15 and others added 6 commits September 14, 2025 19:43

-removed api keys, secrets, client id

0e32505

-removed __pychace__ -added .gitignore

- created SDKPython subdir in labellerr

90572d8

- added utils/client_utils - added video_sampling/pyscene_detect.py

added ffmpeg sampling frames method

9178b37

added SSIM based video keyframe sampling method

03dfa0a

- added gemini method for video sampling

e6e6bba

- add videointelligence module in requirements

ximihoque suggested changes Oct 4, 2025

View reviewed changes

ximihoque and others added 8 commits October 4, 2025 15:56

Review comments

6cc7e52

fixed scripts

18fb31b

Merge branch 'main' of https://github.com/yashsuman15/SDKPython

a6009c6

merging with main

ac7bc84

Merge branch 'main' into feature/LABIMP-7672

2ee6ec9

files restructiing

868101f

Merge branch 'feature/LABIMP-7672' of https://github.com/yashsuman15/…

973cf02

…SDKPython into feature/LABIMP-7672

updated the scripts based on comments

bd1da4c

yashsuman15 commented Oct 6, 2025

View reviewed changes

yashsuman15 added 3 commits October 6, 2025 16:06

added threading to frames downloading

5addbaa

Revert "added threading to frames downloading"

bad6f53

This reverts commit 5addbaa.

minor changes in client_utils

dac6ffb

ximihoque suggested changes Oct 7, 2025

View reviewed changes

remove the file_id argument, only video_path is needed

3fb23e9

yashsuman15 commented Oct 8, 2025

View reviewed changes

yashsuman15 added 8 commits October 8, 2025 18:53

added the cookbook

5a16b51

added the updated client.py added the restructure files

minor update

3758845

minor chages to pyscene storing pattern

194cb26

added video dataset class, minor update video sampling algo

6d118b6

minor update

11b33f8

Refactor video processing methods

e58ba82

minor changes to SDK

67f864e

yashsuman15 and others added 2 commits October 14, 2025 16:58

Resolved conflicts

8330f7b

ximihoque merged commit 551e304 into Labellerr:main Oct 14, 2025
1 of 2 checks passed

		credentials_path: Path to service account JSON key file.
		If None, uses GOOGLE_APPLICATION_CREDENTIALS environment variable

		# https://api.labellerr.com/data/file_data?file_id=c44f38f6-0186-436f-8c2d-ffb50a539c76&include_answers=false&project_id=gabrila_artificial_duck_74237&uuid=1d4c9b58-c6a4-4ca8-9583-b6b6cd25ef12


		class FileMetadataService:

Uh oh!

Conversation

yashsuman15 commented Oct 2, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

yashsuman15 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

yashsuman15 left a comment

Choose a reason for hiding this comment

Uh oh!

ximihoque commented Oct 13, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants