Why Is Frigate Detect FPS Dropping Over Time?
- Detector overloaded (too many cameras/streams)
- Detect fps set too high
- Thermal throttling (TPU/CPU)
Problem Description
Frigate's object detection FPS gradually decreases over hours of operation — detection starts at 5 FPS but drops to 1-2 FPS or lower after several hours, causing missed detections and delayed events. This is typically caused by CPU overload, thermal throttling of the Coral TPU, memory pressure, or constant motion triggering detection on every frame.
Why This Happens in Real Homes
Frigate detect FPS dropping over time means the detector can't keep up with the workload — each camera's frames queue for object detection, and if the detector is overloaded (too many cameras, detect fps set too high, or thermal throttling), latency grows and effective FPS falls. A single Coral or CPU detector has a finite capacity.
Reduce the load: lower the detect fps (5 fps is plenty for most detection), lower the detect resolution, and cut needless detections by tuning motion detection so the detector only runs on real movement. Improve cooling if the Coral or CPU is thermal-throttling under sustained load. If you're simply running more cameras than one detector can handle, add a second Coral or spread the load. Matching the detection workload to the detector's capacity keeps FPS steady.
Symptoms
- Detect FPS drops over time
- Detection slows down
- FPS gradually decreases
- Detector falling behind
- Rising detection latency
- FPS degrades with uptime
- Detection lag grows
- Detector overloaded over time
Recognize these? Here's what usually causes it.
Common Causes
- Detector overloaded (too many cameras/streams)
- Detect fps set too high
- Thermal throttling (TPU/CPU)
- Detect resolution too high
- Too much motion generating detections
- Resource/memory buildup
- CPU fallback under load
- Model too heavy for the hardware
Most fixes happen in the first 3 steps.
Do not optimize from short-run snapshots only.
Tools & Requirements
Step-by-Step Solution
Check CPU and memory usage on the Frigate host
Detection FPS drops when the host runs out of processing power. Run htop or top on the Frigate host and watch CPU usage during active detection. If all cores are at 90%+ consistently: the system cannot keep up with the detection workload. Common causes: too many cameras running detection simultaneously, detect resolution set too high (1080p+ per camera), or other services (Home Assistant, Node-RED) competing for CPU on the same machine.
Reduce the detect resolution per camera
Frigate runs object detection on the detect stream, not the record stream. Higher resolution = more pixels to process per frame = slower detection. In frigate.yml: cameras: your_camera: detect: width: 1280, height: 720 is the recommended maximum. For cameras where you only need to detect people at medium range: 640x480 is sufficient and uses 75% less processing power. The record stream can remain at full resolution — detection and recording are independent.
Verify Coral TPU is connected and processing
Without a Coral TPU, detection runs on CPU — which is 10-50x slower. Check Frigate logs for 'Detector started' messages. If it says 'cpu' instead of 'edgetpu': the Coral is not detected. For USB Coral: check lsusb for 'Global Unichip Corp' (device ID 18d1:9302 or 1a6e:089a). For PCIe Coral: check ls /dev/apex_0. If the device exists but Frigate does not see it: check Docker device passthrough. In docker-compose.yml: devices: - /dev/bus/usb:/dev/bus/usb (USB) or - /dev/apex_0:/dev/apex_0 (PCIe).
Limit detection to motion-only frames
Frigate only runs detection when motion is detected — but if a camera has constant motion (trees blowing, busy street, flickering lights): detection runs on every frame, overloading the system. Use motion masks to block areas with constant movement: cameras: your_camera: motion: mask: [coordinates]. Also adjust motion threshold and contour area to reduce false motion triggers: motion: threshold: 30 (default 25, higher = less sensitive), contour_area: 30 (default 20, higher = ignores smaller motion regions).
Monitor detection stats over time in the Frigate UI
The Frigate web UI shows real-time detection statistics: detection FPS, inference time (ms per frame), and CPU/GPU usage per detector. Go to System in the Frigate UI. A healthy Coral TPU shows inference times of 8-15ms. If inference time creeps up over hours: the Coral may be thermal throttling (check if the TPU is in an enclosed space without airflow). If detection FPS drops but inference time stays low: the bottleneck is the FFmpeg decode stage, not detection — reduce the number of cameras or lower detect FPS: detect: fps: 5 (default 5, lower to 3 for non-critical cameras).
Quick Solutions
Still having issues? This is usually the deeper cause below.
If the sensor still misses events after repositioning, check whether a scheduled 'home' or 'away' mode is overriding the sensitivity setting silently.
Time-based performance charts are essential for diagnosing gradual degradation.
This issue almost always looks more complex than it is — the majority of cases trace back to a single setting, a stale credential, or a default that shipped wrong.
- Detector overloaded (too many cameras/streams)
- Detect fps set too high
- Thermal throttling (TPU/CPU)
- Detect resolution too high
- Too much motion generating detections
Before you go — try one of these (they fix most cases).
Official Manufacturer Manual
Frigate provides official product documentation through their online manual rather than downloadable PDF. Access setup guides, troubleshooting steps, and product specifications for your Frigate Detect FPS Degradation.
Source: docs.frigate.video
Need More Help? Frigate Support
Note: The contact information below connects you directly to Frigate's official customer support team, not Trunetto. They can help with warranty claims, device replacements, and advanced technical issues.
Accessories owners commonly pair with Frigate Detect FPS Degradation.
SANDISK 64GB Extreme microSDXC UHS-I Memory Card with Ada...
SANDISK 128GB Ultra microSDXC UHS-I Memory Card with Adap...
SANDISK 64GB Ultra microSDXC UHS-I Memory Card with Adapt...
As an Amazon Associate we earn from qualifying purchases.