Object area debugging and improvements (#16432)

* add ability to specify min and max area as percentages

* debug draw area and ratio

* docs

* update for best percentage
This commit is contained in:
Josh Hawkins 2025-02-09 15:48:23 -06:00 committed by GitHub
parent 83beacf84a
commit c8cec63cb9
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
8 changed files with 352 additions and 21 deletions

View File

@ -34,7 +34,7 @@ False positives can also be reduced by filtering a detection based on its shape.
### Object Area
`min_area` and `max_area` filter on the area of an objects bounding box in pixels and can be used to reduce false positives that are outside the range of expected sizes. For example when a leaf is detected as a dog or when a large tree is detected as a person, these can be reduced by adding a `min_area` / `max_area` filter.
`min_area` and `max_area` filter on the area of an objects bounding box and can be used to reduce false positives that are outside the range of expected sizes. For example when a leaf is detected as a dog or when a large tree is detected as a person, these can be reduced by adding a `min_area` / `max_area` filter. These values can either be in pixels or as a percentage of the frame (for example, 0.12 represents 12% of the frame).
### Object Proportions

View File

@ -312,9 +312,11 @@ objects:
# Optional: filters to reduce false positives for specific object types
filters:
person:
# Optional: minimum width*height of the bounding box for the detected object (default: 0)
# Optional: minimum size of the bounding box for the detected object (default: 0).
# Can be specified as an integer for width*height in pixels or as a decimal representing the percentage of the frame (0.000001 to 0.99).
min_area: 5000
# Optional: maximum width*height of the bounding box for the detected object (default: 24000000)
# Optional: maximum size of the bounding box for the detected object (default: 24000000).
# Can be specified as an integer for width*height in pixels or as a decimal representing the percentage of the frame (0.000001 to 0.99).
max_area: 100000
# Optional: minimum width/height of the bounding box for the detected object (default: 0)
min_ratio: 0.5
@ -559,7 +561,7 @@ genai:
# Optional: Restream configuration
# Uses https://github.com/AlexxIT/go2rtc (v1.9.2)
# NOTE: The default go2rtc API port (1984) must be used,
# changing this port for the integrated go2rtc instance is not supported.
# changing this port for the integrated go2rtc instance is not supported.
go2rtc:
# Optional: Live stream configuration for WebUI.

View File

@ -11,11 +11,13 @@ DEFAULT_TRACKED_OBJECTS = ["person"]
class FilterConfig(FrigateBaseModel):
min_area: int = Field(
default=0, title="Minimum area of bounding box for object to be counted."
min_area: Union[int, float] = Field(
default=0,
title="Minimum area of bounding box for object to be counted. Can be pixels (int) or percentage (float between 0.000001 and 0.99).",
)
max_area: int = Field(
default=24000000, title="Maximum area of bounding box for object to be counted."
max_area: Union[int, float] = Field(
default=24000000,
title="Maximum area of bounding box for object to be counted. Can be pixels (int) or percentage (float between 0.000001 and 0.99).",
)
min_ratio: float = Field(
default=0,

View File

@ -29,6 +29,7 @@ from frigate.util.builtin import (
)
from frigate.util.config import (
StreamInfoRetriever,
convert_area_to_pixels,
find_config_file,
get_relative_coordinates,
migrate_frigate_config,
@ -148,6 +149,13 @@ class RuntimeFilterConfig(FilterConfig):
if mask is not None:
config["mask"] = create_mask(frame_shape, mask)
# Convert min_area and max_area to pixels if they're percentages
if "min_area" in config:
config["min_area"] = convert_area_to_pixels(config["min_area"], frame_shape)
if "max_area" in config:
config["max_area"] = convert_area_to_pixels(config["max_area"], frame_shape)
super().__init__(**config)
def dict(self, **kwargs):

View File

@ -347,6 +347,36 @@ def get_relative_coordinates(
return mask
def convert_area_to_pixels(
area_value: Union[int, float], frame_shape: tuple[int, int]
) -> int:
"""
Convert area specification to pixels.
Args:
area_value: Area value (pixels or percentage)
frame_shape: Tuple of (height, width) for the frame
Returns:
Area in pixels
"""
# If already an integer, assume it's in pixels
if isinstance(area_value, int):
return area_value
# Check if it's a percentage
if isinstance(area_value, float):
if 0.000001 <= area_value <= 0.99:
frame_area = frame_shape[0] * frame_shape[1]
return max(1, int(frame_area * area_value))
else:
raise ValueError(
f"Percentage must be between 0.000001 and 0.99, got {area_value}"
)
raise TypeError(f"Unexpected type for area: {type(area_value)}")
class StreamInfoRetriever:
def __init__(self) -> None:
self.stream_cache: dict[str, tuple[int, int]] = {}

View File

@ -0,0 +1,177 @@
import React, { useState, useRef, useCallback, useMemo } from "react";
import { Stage, Layer, Rect } from "react-konva";
import { KonvaEventObject } from "konva/lib/Node";
import {
Popover,
PopoverContent,
PopoverTrigger,
} from "@/components/ui/popover";
import Konva from "konva";
import { useResizeObserver } from "@/hooks/resize-observer";
type DebugDrawingLayerProps = {
containerRef: React.RefObject<HTMLDivElement>;
cameraWidth: number;
cameraHeight: number;
};
function DebugDrawingLayer({
containerRef,
cameraWidth,
cameraHeight,
}: DebugDrawingLayerProps) {
const [rectangle, setRectangle] = useState<{
x: number;
y: number;
width: number;
height: number;
} | null>(null);
const [isDrawing, setIsDrawing] = useState(false);
const [showPopover, setShowPopover] = useState(false);
const stageRef = useRef<Konva.Stage>(null);
const [{ width: containerWidth }] = useResizeObserver(containerRef);
const imageSize = useMemo(() => {
const aspectRatio = cameraWidth / cameraHeight;
const imageWidth = containerWidth;
const imageHeight = imageWidth / aspectRatio;
return { width: imageWidth, height: imageHeight };
}, [containerWidth, cameraWidth, cameraHeight]);
const handleMouseDown = (e: KonvaEventObject<MouseEvent>) => {
const pos = e.target.getStage()?.getPointerPosition();
if (pos) {
setIsDrawing(true);
setRectangle({ x: pos.x, y: pos.y, width: 0, height: 0 });
}
};
const handleMouseMove = (e: KonvaEventObject<MouseEvent>) => {
if (!isDrawing) return;
const pos = e.target.getStage()?.getPointerPosition();
if (pos && rectangle) {
setRectangle({
...rectangle,
width: pos.x - rectangle.x,
height: pos.y - rectangle.y,
});
}
};
const handleMouseUp = () => {
setIsDrawing(false);
if (rectangle) {
setShowPopover(true);
}
};
const convertToRealCoordinates = useCallback(
(x: number, y: number, width: number, height: number) => {
const scaleX = cameraWidth / imageSize.width;
const scaleY = cameraHeight / imageSize.height;
return {
x: x * scaleX,
y: y * scaleY,
width: width * scaleX,
height: height * scaleY,
};
},
[cameraWidth, cameraHeight, imageSize.width, imageSize.height],
);
const calculateArea = useCallback(() => {
if (!rectangle) return 0;
const { width, height } = convertToRealCoordinates(
0,
0,
Math.abs(rectangle.width),
Math.abs(rectangle.height),
);
return width * height;
}, [rectangle, convertToRealCoordinates]);
const calculateAreaPercentage = useCallback(() => {
if (!rectangle) return 0;
const { width, height } = convertToRealCoordinates(
0,
0,
Math.abs(rectangle.width),
Math.abs(rectangle.height),
);
return (width * height) / (cameraWidth * cameraHeight);
}, [rectangle, convertToRealCoordinates, cameraWidth, cameraHeight]);
const calculateRatio = useCallback(() => {
if (!rectangle) return 0;
const { width, height } = convertToRealCoordinates(
0,
0,
Math.abs(rectangle.width),
Math.abs(rectangle.height),
);
return width / height;
}, [rectangle, convertToRealCoordinates]);
return (
<div className="absolute inset-0 cursor-crosshair">
<Stage
width={imageSize.width}
height={imageSize.height}
onMouseDown={handleMouseDown}
onMouseMove={handleMouseMove}
onMouseUp={handleMouseUp}
ref={stageRef}
>
<Layer>
{rectangle && (
<Rect
x={rectangle.x}
y={rectangle.y}
width={rectangle.width}
height={rectangle.height}
stroke="white"
strokeWidth={4}
/>
)}
</Layer>
</Stage>
{showPopover && rectangle && (
<Popover open={showPopover} onOpenChange={setShowPopover}>
<PopoverTrigger asChild>
<div
style={{
position: "absolute",
left: `${rectangle.x + rectangle.width / 2}px`,
top: `${rectangle.y + rectangle.height / 2}px`,
}}
/>
</PopoverTrigger>
<PopoverContent className="w-auto p-5 text-center">
<div className="flex flex-col gap-2">
<div className="flex flex-col text-primary">
Area:{" "}
<span className="text-sm text-primary-variant">
px: {calculateArea().toFixed(0)}
</span>
<span className="text-sm text-primary-variant">
%: {calculateAreaPercentage().toFixed(4)}
</span>
</div>
<div className="flex flex-col text-primary">
Ratio:{" "}
<span className="text-sm text-primary-variant">
{" "}
{calculateRatio().toFixed(2)}
</span>
</div>
</div>
</PopoverContent>
</Popover>
)}
</div>
);
}
export default DebugDrawingLayer;

View File

@ -490,12 +490,27 @@ export default function ObjectLifecycle({
Area
</p>
{Array.isArray(item.data.box) &&
item.data.box.length >= 4
? Math.round(
detectArea *
(item.data.box[2] * item.data.box[3]),
)
: "N/A"}
item.data.box.length >= 4 ? (
<>
<div className="flex flex-col text-xs">
px:{" "}
{Math.round(
detectArea *
(item.data.box[2] * item.data.box[3]),
)}
</div>
<div className="flex flex-col text-xs">
%:{" "}
{(
(detectArea *
(item.data.box[2] * item.data.box[3])) /
detectArea
).toFixed(4)}
</div>
</>
) : (
"N/A"
)}
</div>
</div>
</div>

View File

@ -1,4 +1,4 @@
import { useCallback, useEffect, useMemo } from "react";
import { useCallback, useEffect, useMemo, useRef, useState } from "react";
import ActivityIndicator from "@/components/indicators/activity-indicator";
import AutoUpdatingCameraImage from "@/components/camera/AutoUpdatingCameraImage";
import { CameraConfig, FrigateConfig } from "@/types/frigateConfig";
@ -23,6 +23,9 @@ import { getIconForLabel } from "@/utils/iconUtil";
import { capitalizeFirstLetter } from "@/utils/stringUtil";
import { LuExternalLink, LuInfo } from "react-icons/lu";
import { Link } from "react-router-dom";
import DebugDrawingLayer from "@/components/overlay/DebugDrawingLayer";
import { Separator } from "@/components/ui/separator";
import { isDesktop } from "react-device-detect";
type ObjectSettingsViewProps = {
selectedCamera?: string;
@ -37,6 +40,8 @@ export default function ObjectSettingsView({
}: ObjectSettingsViewProps) {
const { data: config } = useSWR<FrigateConfig>("config");
const containerRef = useRef<HTMLDivElement>(null);
const DEBUG_OPTIONS = [
{
param: "bbox",
@ -130,6 +135,12 @@ export default function ObjectSettingsView({
[options, setOptions],
);
const [debugDraw, setDebugDraw] = useState(false);
useEffect(() => {
setDebugDraw(false);
}, [selectedCamera]);
const cameraConfig = useMemo(() => {
if (config && selectedCamera) {
return config.cameras[selectedCamera];
@ -234,7 +245,7 @@ export default function ObjectSettingsView({
<span className="sr-only">Info</span>
</div>
</PopoverTrigger>
<PopoverContent className="w-80">
<PopoverContent className="w-80 text-sm">
{info}
</PopoverContent>
</Popover>
@ -256,18 +267,74 @@ export default function ObjectSettingsView({
</div>
))}
</div>
{isDesktop && (
<>
<Separator className="my-2" />
<div className="flex w-full flex-row items-center justify-between">
<div className="mb-2 flex flex-col">
<div className="flex items-center gap-2">
<Label
className="mb-0 cursor-pointer capitalize text-primary"
htmlFor="debugdraw"
>
Object Shape Filter Drawing
</Label>
<Popover>
<PopoverTrigger asChild>
<div className="cursor-pointer p-0">
<LuInfo className="size-4" />
<span className="sr-only">Info</span>
</div>
</PopoverTrigger>
<PopoverContent className="w-80 text-sm">
Enable this option to draw a rectangle on the
camera image to show its area and ratio. These
values can then be used to set object shape filter
parameters in your config.
<div className="mt-2 flex items-center text-primary">
<Link
to="https://docs.frigate.video/configuration/object_filters#object-shape"
target="_blank"
rel="noopener noreferrer"
className="inline"
>
Read the documentation{" "}
<LuExternalLink className="ml-2 inline-flex size-3" />
</Link>
</div>
</PopoverContent>
</Popover>
</div>
<div className="mt-1 text-xs text-muted-foreground">
Draw a rectangle on the image to view area and ratio
details
</div>
</div>
<Switch
key={`$draw-${selectedCamera}`}
className="ml-1"
id="debug_draw"
checked={debugDraw}
onCheckedChange={(isChecked) => {
setDebugDraw(isChecked);
}}
/>
</div>
</>
)}
</div>
</div>
</TabsContent>
<TabsContent value="objectlist">
{ObjectList(memoizedObjects)}
<ObjectList cameraConfig={cameraConfig} objects={memoizedObjects} />
</TabsContent>
</Tabs>
</div>
{cameraConfig ? (
<div className="flex md:h-dvh md:max-h-full md:w-7/12 md:grow">
<div className="size-full min-h-10">
<div ref={containerRef} className="relative size-full min-h-10">
<AutoUpdatingCameraImage
camera={cameraConfig.name}
searchParams={searchParams}
@ -275,6 +342,13 @@ export default function ObjectSettingsView({
className="size-full"
cameraClasses="relative w-full h-full flex flex-col justify-start"
/>
{debugDraw && (
<DebugDrawingLayer
containerRef={containerRef}
cameraWidth={cameraConfig.detect.width}
cameraHeight={cameraConfig.detect.height}
/>
)}
</div>
</div>
) : (
@ -284,7 +358,12 @@ export default function ObjectSettingsView({
);
}
function ObjectList(objects?: ObjectType[]) {
type ObjectListProps = {
cameraConfig: CameraConfig;
objects?: ObjectType[];
};
function ObjectList({ cameraConfig, objects }: ObjectListProps) {
const { data: config } = useSWR<FrigateConfig>("config");
const colormap = useMemo(() => {
@ -326,7 +405,7 @@ function ObjectList(objects?: ObjectType[]) {
{capitalizeFirstLetter(obj.label.replaceAll("_", " "))}
</div>
</div>
<div className="flex w-8/12 flex-row items-end justify-end">
<div className="flex w-8/12 flex-row items-center justify-end">
<div className="text-md mr-2 w-1/3">
<div className="flex flex-col items-end justify-end">
<p className="mb-1.5 text-sm text-primary-variant">
@ -351,7 +430,25 @@ function ObjectList(objects?: ObjectType[]) {
<p className="mb-1.5 text-sm text-primary-variant">
Area
</p>
{obj.area ? obj.area.toString() : "-"}
{obj.area ? (
<>
<div className="text-xs">
px: {obj.area.toString()}
</div>
<div className="text-xs">
%:{" "}
{(
obj.area /
(cameraConfig.detect.width *
cameraConfig.detect.height)
)
.toFixed(4)
.toString()}
</div>
</>
) : (
"-"
)}
</div>
</div>
</div>