Bunch of fixes and new features #250
Merged
Conversation
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
instead of project_page.py - Add connection between login page and project page - Improve the user creation functions to use better DB queries - Add connection between different user pages - Change theme to only use 'Light' theme because the `data_table` would have weird colors if using any other theme - Commented out every line of `st.set_page_config()` - TBC on user_management_page and user_info page
Dashboard, CreateNewUser, and EditUser - Use create_new_user page to directly work with EditUser by checking whether session_state.current_user exists
page with User Session Log table - New: Add "Switch User" navigation in deployment section to allow switching to other user even during deployment
- Also fix the `session_state.deployment_conf` to use `on_change` or `on_click` callbacks to update the `deploy_conf` on any widget change - Fix `update_logout_session_log()`
either classification or segmentation, if labelmap file is provided - Add labelmap generation during training - Change uploaded type for keras models to only accept archive, similar to TFOD - Add checking for output layer nodes to be same with number of classes in uploaded labelmap file or the generated labelmap - Various fixes to deployment page to save frames properly - Fix exporting for evaluation if the exported files were deleted after training - Change models_page to directly to go training_config page if attached_model is changed, to avoid issues with segmentation model parameters - Also reset session_state.trainer if user decided to go back to change any existing info after entering training page - Fix augmentation page to work for uploaded image - Improved training_config to use `param_dict.get()` to get training_param - Change Training.training_path from property to `get_paths()` method - Use `keras-model.h5` as model filename to avoid issues when user decided to change model name
- Send the user to database_setup page if no secrets.toml file or if database connection failed - Allow user info update without password update - Reset all session_states and logout if user decides to delete himself - Change the user role of first user created must be Administrator - Add all latest package requirements to 'requirements_no_hash.txt' - Also fix the TFOD installation
- Use existing database of name 'integrated_vision_inspection_system' if already exists - Show clearer instructions to the users in the database_setup form
- Use pycocotools-windows for Windows installation - Fix most st.cache decorated functions by using allow_output_mutation=True, not sure why there was no such error previously with st.cache - Use APP_DATABASE_NAME to get from env variable or default to "integrated_vision_inspection_system" to create this db name for our app - Fix visualization of segment_inference_pipeline for deployment - Improve helper.get_directory_name() to remove unwanted symbols, also change Project.query_annotations to create image_paths the same way - Fix editor_management's TAGNAMES to use PolygonLabels for Semantic - Test and fix uploading and training of segmentation models
in settings page
- Add more instructions in deployment_page
- Add more instructions in deployment_page - Feat: add publish_frame config to publish output_frame through MQTT
…ated-Vision-Inspection-System into misc_fix_and_feats
to work for all pretrained_models - Add different preprocess_input function for pretrained classification models - Fix custom_train_test_split() to work properly for stratify with val + test set - Change xml_to_df() to only use cache in augmentation_config page
- Fix PostgreSQL port to always use 5432 in docker-compose
- Add docker-compose.cpu.yml file for without GPU support - Change '.env' file as optional for docker-compose - Set user to be 'postgres' as it is the default superuser
- Set the maximum length of a name to be 21 (in check_if_field_empty()) - Fix `find_tfod_metric()` to take the entire `Loss/...' metric name
- To handle paths with space characters, in functions - Removed session_state.archive_success - Replace st.experimental_memo with st.cache to be able to clear their cache
in deployment page - Fix updating existing user info without any password input - Add selectbox in User Management Page to easily change user role
- Rearrange MQTT options to appear before clicking "Deploy Model"
- Add timezone to get_result_fn in deployment - Fix the run_command() to correctly spaced in trainer
at different views
- Use MQTT to send the current view from dobot script to the deployment_page to be able to check the labels - Add labels for different views for P2/143 machine part - TO TEST for P2/143 training and running
move_for_p2_143()
- Store NG images in a specific 'NG-images' directory
Updated Dobot API
- Add/Remove a few attributes in DeploymentConfig: use_multi_cam, num_cameras, camera_titles, camera_types, camera_sources, camera_keys - camera_sources attribute is used to combine both USB camera ports and IP camera addresses, make things easier - Change MQTTTopics.publish_frame to be List[str] to publish frame to different topics for different cameras - Video recording is also saved into separate files for each camera - Label checking at different views is also possible by comparing received 'view' message with the 'camera_titles' entered by the user
…ated-Vision-Inspection-System into misc_fix_and_feats
- get labels from classes in main.py
- Use frontslash '/' in the camera title to separate multiple views for a single camera
…ated-Vision-Inspection-System into misc_fix_and_feats
- Add a spinner message for the "Move Dobot" button in deployment
- Use os.kill(pid, signal) to kill the subprocess' pid
- By changing checkpoint_every_n to save ckpt at every 100 steps instead of 1000 to be able to continue from any trained steps of multiple of 100 Also fix deployment page issue with updating number of cameras too fast
- Fix Docker permission issues when saving to existing CSV files during database creation
- Fix for both Dockerfile and Dockerfile.cpu
- Change Docker postgres local port to 6543 - Fix restarting model sequence properly for pretrained models in scrape_setup_model_details() - Updated streamlit 1.3.0 -> 1.3.1
- by adding the "myuser" to "video" usergroup - Change list_available_cameras() to check for at least 5 ports
- to avoid permission issues if downloading to cwd in Docker - Fix the captured_images path in new_dataset.py if using Docker container
- Change create_tf_dataset() to use small shuffle buffer size without cache to avoid slow and memory issues - Change new_dataset page to use the proper "cam_key" for IP Camera to reset properly
- Fix: create all required directories in path_desc.py to avoid issues
- By reset_multi_camera_config() at the right places, and also update_camera_sources_from_types() when needed - Fix error checking to check that number of USB cameras specified is not more than the number of working ports - Add a "Reset all cameras" button to use in case any error occurs
- For CSV, only save if the CSV file is not exists, or if the new scraped data is not the same as existing CSV data - Same thing for database models table pretrained model rows
- by allowing user to configure how many ports to check exactly - Refactor create_csv_file_and_writer() to check whether the CSV file exists within the function
- Update config.toml to let new Streamlit work properly - remove add_report_ctx() as it is not available and not needed anymore
for image classification deployment type
This was
linked to
issues
Jan 24, 2022
Closed
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Essentially a lot of bug fixes and then updates to the main app.py script, and also deployment_page.py. For more details, refer to the commit messages as they explained everything in detail.