AI Photos Feature Guide
The AI Photos feature performs Face Recognition and Smart Recognition locally on your Fygo NAS for your photos and videos, enabling features such as People albums, AI photo search, smart categories, and enhanced video recognition to help you manage your photo library efficiently.
1. Installation Guide
1.1. Hardware Requirements
Since the AI Photos feature runs models locally, there are minimum hardware requirements for the NAS.
- Minimum 2 GB of RAM
- At least 20 GB of free storage space at the model installation location. The model and related files will be stored on the same volume.
1.2. Download and Install Models
Face Recognition and Smart Recognition require separate model downloads. The app will automatically detect the device's hardware configuration and recommend suitable models.
-
Face Recognition: Only the "Standard Model" is available
- If your photo library contains children or people with similar facial features, you can adjust the Face Similarity Preference for better differentiation
-
Smart Recognition: "Basic Model" or "Enhanced Model" available depending on hardware
- Basic Model: Minimum 2 GB RAM
- Enhanced Model: Minimum 8 GB RAM, requires more storage space and computing resources, but provides better accuracy and understanding
- Japanese-Optimized Model: Supports Japanese search and is suitable for users in Japan or scenarios where Japanese is the primary language. It performs better when searching in Japanese.
You can download and install models in 2 ways:
-
Select model file (.zip)
-
You can pre-download the model file to the NAS, then open Fygo Photos and select the model file to extract and install. Do not extract the model file manually; the Fygo Photos app will verify and extract it automatically.
-
Download methods:
-
Copy the following magnet links and open them in Download to download to the NAS
-
FygoOS ARM full model pack:
-
FygoOS x86 full model pack:
-
ARM individual model packages:
-
Face Recognition, Standard Model V1.0.1
-
Smart Recognition, Base Model
-
Smart Recognition, Enhanced Model
-
Smart Recognition, Japanese-Optimized Model
-
-
x86 individual model packages:
-
Face Recognition, Standard Model V1.0.1
-
Smart Recognition, Base Model
-
Smart Recognition, Enhanced Model
-
Smart Recognition, Japanese-Optimized Model
-
-
-
-
Direct download and install
- Download speed may be slow due to concurrent downloads from multiple users, requiring more time
If you encounter errors during download/extraction and installation, the following situations may apply:
- Download and install
Network error: Check if the network connection is successful
Insufficient storage: Check if the selected model installation volume has at least 20 GB of free space
No storage permission: Check if you have the required permissions
For "exit status 1" errors when downloading models, please manually troubleshoot:
Check if initialization dependencies are installed
apt list | grep python3.11-venv
apt list | grep python3-pip
If not installed, run the commands manually and check for errors
apt install python3.11-venv && apt install python3-pip
- Select model file to install
- Installation failed, invalid file. Please verify the model file was downloaded from an official source.
- Installation failed, not a face recognition/smart recognition model file. Please check if the selected file is correct.
- Installation failed, cannot install this model, please check hardware. The Enhanced Model for Smart Recognition requires 8 GB of RAM. If RAM is insufficient, please adjust hardware or use the Basic Model.
1.3. AI Hardware Acceleration
Fygo Photos can enable AI hardware acceleration to reduce CPU usage and improve AI task processing speed. It supports NVIDIA GPUs as well as Intel / AMD / ARM GPUs or NPUs.
If GPU invocation fails, the following situations may apply:
- NVIDIA GPU
- Missing GPU driver: Please install the NVIDIA GPU driver from the App Center. Restart the NAS after installing the GPU driver.
- Insufficient VRAM: Check if other non-Photos tasks are running simultaneously. You can close other non-Photos application tasks.
- Intel GPU
- GPU too busy: Check if other non-Photos tasks are running simultaneously. You can close other non-Photos application tasks.
- Older GPUs may not be supported.
2. Task Execution
Tasks can be triggered in the following ways:
- After installing a model, tasks will automatically starand execute
- After enabling the feature, tasks will automatically execute some time after uploading photos or videos
- Manually click Process Unrecognized Photos to check for unprocessed photos or videos and add recognition tasks
- Manually click Reset Recognition to delete existing people or category data and re-recognize all data
- If Enhanced Video Recognition is also enabled, clicking Reset Smart Recognition will also reset enhanced video recognition data and execute tasks
If task execution fails, the following reasons may apply:
Smart Recognition: Algorithm startup timeout
If Face Recognition works normally but Smart Recognition reports "Algorithm startup timeout", it is caused by GPU incompatibility. Temporarily disable GPU and retry. Some GPUs require additional startup time.
Smart Recognition: Service error
Caused by GPU compatibility issues. Some GPUs may throw service exceptions mid-execution. Temporarily disable GPU and retry. This issue is currently being investigated.
Face Recognition: Image does not exist
In abnormal scenarios (e.g., disk failure, external storage disconnection, file deletion), please first check storage, external storage, and file permissions. The thumbnail's original path may not have been updated in time. Go to Files to manually trigger a scan, then re-execute the face recognition task.
Face/Smart Recognition: TRPC read timeout
This may be caused by CPU/IO being busy. It is recommended to retry the task when resources are available.
3. AI Photos Feature Guide
3.1. Face Recognition Advanced Settings
Face Recognition has 3 key parameters for you to configure.
-
Face Detection Confidence Threshold
This value affects the accuracy of determining whether a face exists in a photo or video during recognition.
-
Increasing this value raises the requirements and accuracy for face detection, but may also filter out photos or videos that potentially contain faces.
-
Decreasing this value may result in incorrect face recognition.
- Not recommended to modify.
- Reset Face Recognition is required after modification.
-
-
Minimum Photos to Generate a Person
This value affects the number of photos or videos required to aggregate into a People album. If you find too many people albums are generated and want to reduce irrelevant ones, increase this value.
- Reset Face Recognition is not required after modification. Only re-aggregation is needed, which is a quick process.
-
Person Similarity Difference Threshold
This value affects the similarity requirement for aggregating into a People album.
-
Decreasing this value raises the similarity requirement for grouping as the same person, making people more accurate. However, the same person may be split into multiple People albums.
-
Increasing this value may result in different people's photos being grouped into the same People album.
Adjustment notes:
-
If your photo library does not contain many people and the differentiation requirement is not high, try Loose Aggregation to group more photos under the same person.
-
Reset Face Recognition is not required after modification. Only re-aggregation is needed, which is a quick process.
-
3.2. People
After Face Recognition completes, People albums will be automatically generated based on the parameters in Advanced Settings. You can view them in the People section of the sidebar.
After people are aggregated, you can perform the following operations:
- Name a person for easier identification, searching, and filtering
- View the person in Preview
- Input or filter by person in Search & Filter
- If the same person is split into multiple People albums, you can Merge People
- Show/Hide a person: When hidden, the person will not appear in Preview, People list, or Search
- If you find photos of other people in a person's album, you can Remove this photo from the person
- In Albums, you can set a person as a condition for automatic addition
If a person's photos in the library do not meet the minimum photo count to generate a People album, you can still manually add faces and create a person in Preview.
You can also add other photos with similar faces (photos not yet in the person's album) to the same person.
3.3. AI Search
Toggle the AI Search switch to enable AI photo search mode. AI Search allows you to enter a description to search for photos and videos in your library.
You can try entering short descriptions to search for photos, such as:
- Holding a child / Child sleeping
- Walking by the (beach/street/...)
- Eating (noodles/pizza/...)
- Smiling happily
- ...
If your library contains many videos, try enabling Enhanced Video Recognition for better results.
3.4. Smart Categories
After Smart Recognition is complete, if you have a sufficient number of photos or videos, smart categories will be automatically generated. You can view them in the Smart Categories section of the sidebar.
3.5. Enhanced Video Recognition
Enabling Smart Recognition performs basic recognition on videos by default. Enabling Enhanced Video Recognition performs enhanced recognition on videos. After enhanced recognition is complete, more videos can be found through search. If your library contains many videos, it is recommended to enable Enhanced Video Recognition for a better search experience.
Note:
- Enhanced Video Recognition only processes .mp4 and .mov format videos
- Enhanced Video Recognition consumes more computing resources and takes a long time!
- It is recommended to complete Face Recognition and Smart Recognition first, then run Enhanced Video Recognition separately