-
Notifications
You must be signed in to change notification settings - Fork 0
3_FAQ
Common answers and troubleshooting tips for DRerio LogAI users.
A: Since v1.6 the project creation wizard is enabled by default. If you disabled it previously, add the following to config.local.yaml:
ui_features:
use_wizard_for_project_creation: trueRestart the application and the 5-step wizard will appear when you click Create Project.
- Lighting & Contrast: Ensure the arena background contrasts with the animal. Avoid reflections or strong shadows.
- Processing Area: Confirm the arena polygon fully covers the region where the animal moves. With the new clamping logic, the snap indicator stays inside the arena — if you cannot place a point, the arena is probably too small.
- Confidence / NMS: Lower the detector confidence threshold (e.g., from 0.85 to 0.65) or tweak NMS in the advanced configuration tab.
- Single-subject mode: Calibration and diagnostics force single-subject tracking. The overlay shows the active mode; make sure the correct mode is being used for your experiment.
A: MP4 (H.264) at 25-30 fps offers the best balance between size and quality. Higher bitrates improve detection, but ensure your storage and GPU can keep up. Other formats (AVI, MOV) work as long as OpenCV can decode them.
A: Draw the arena/ROIs once, then use 💾 Salvar Zonas Atuais to store the layout in templates/. Later projects can:
- Select a saved template from Templates salvos and click Aplicar.
- Import external JSON files via 📂 Importar e Aplicar Arquivo... (applies immediately and saves to the library).
Templates include arena, ROI polygons, names, and colors. The wizard also reuses templates when you import Parquet files containing zones.
A: Yes. Calibration translates pixels into centimeters. Without it, distances and velocities are expressed in pixels. Use the calibration dialog (or wizard Step 4 for live projects) to record real-world dimensions. Once the calibration is saved, Recorder adds x_cm/y_cm columns and reports show metrics in centimeters.
A: Yes. Draw the arena corners in the zone editor; DRerio LogAI computes a homography that warps coordinates into a normalized top-down plane. Metrics and ROI checks operate in this warped space, so moderate perspective skew is fine.
A: The GUI automatically refreshes after each processing batch. If you still see an empty list:
- Check the status bar for errors while processing.
- Click Atualizar in the Reports tab to rescan the project folder.
- Confirm that the
<video>_results/folder exists and that3_CoordMovimento_*.parquetwas written (absence indicates processing failed earlier).
- Sharp Turns: Counts how often angular velocity exceeds the configured threshold (default 90°/s).
- Speed bursts / inactivity: Duration and counts for periods above/below configurable thresholds.
- Intra-ROI metrics: Distance, average speed, freezing, and transitions calculated solely within each ROI.
- Social proximity: Percentage of time individuals remain inside dynamically computed proximity clusters (requires multi-animal tracking and calibration).
See docs/reference/operational_reference.md Section 5 for formulas and exact definitions.
A: Each video produces a <video_name>_results/ folder containing:
-
1_ProcessingArea_*.parquet(arena),2_AreasOfInterest_*.parquet(ROIs),3_CoordMovimento_*.parquet(trajectory) - Optional MP4 with overlays if recording was enabled
-
{video}_summary.xlsx,{video}_report.docx, plus optional CSV/Parquet exports
Project-wide exports created from the Relatórios tab are saved wherever you point the file dialog. The wizard also keeps metadata in project_config.json and config_snapshot.yaml for reproducibility.
Open an issue on GitHub describing your environment (OS, Python version, GPU), attach logs (logs/ folder if present) and, if possible, include the wizard summary exported as JSON. Contributions and pull requests are welcome — read CONTRIBUTING.md for guidelines.
A: Reports now include enhanced velocity statistics:
| Metric | Description |
|---|---|
| Mean Speed (cm/s) | Average swimming velocity |
| Max Speed (cm/s) | Maximum instantaneous velocity |
| Median Speed (cm/s) | Median velocity (robust to outliers) |
| Std Speed (cm/s) | Standard deviation of velocity |
Max Speed was added in v3.2 to help identify burst swimming behavior.
A: Geotaxis analysis measures vertical position preference in lateral-view aquariums:
- Geotaxis Zona 1 - Fundo (%) - Time in bottom zone (zone 0 internally)
- Geotaxis Zona 2 (%) - Time in middle zone
- Geotaxis Zona 3 (%) - Time in top zone (if 3 zones configured)
Zone names are 1-indexed for user display (Zona 1, 2, 3...) but stored as 0-indexed internally (zone_0, zone_1, zone_2...).
Configuration: Set geotaxis_num_zones in behavioral analysis settings (default: 3).
A: This was fixed in v3.2 (Dec 2025). If you're seeing empty cells:
-
Update to v3.2+ - Earlier versions had a bug where
behavioral_configwasn't properly passed during summary generation. - Re-process summaries - Go to Reports tab and regenerate the unified report.
Unified reports now include:
- Subject columns (group, subject, day, experiment_id) appearing first
- All geotaxis zone percentages properly populated
- Proper column names with units