videoanalysis: include ffmpeg frame grabber

This commit is contained in:
Koushik Dutta
2023-03-21 23:00:08 -07:00
parent 0dc75bf737
commit beb53c672c
4 changed files with 1064 additions and 296 deletions

File diff suppressed because it is too large Load Diff

View File

@@ -36,13 +36,17 @@
"type": "API",
"interfaces": [
"Settings",
"MixinProvider"
"MixinProvider",
"DeviceProvider"
],
"realfs": true,
"pluginDependencies": [
"@scrypted/python-codecs"
]
},
"optionalDependencies": {
"sharp": "^0.31.3"
},
"dependencies": {
"@scrypted/common": "file:../../common",
"@scrypted/sdk": "file:../../sdk",
@@ -55,6 +59,6 @@
"@types/lodash": "^4.14.175",
"@types/node": "^14.17.11",
"@types/semver": "^7.3.13",
"ts-node": "^10.9.1"
"@types/sharp": "^0.31.1"
}
}

View File

@@ -0,0 +1,196 @@
import { Deferred } from "@scrypted/common/src/deferred";
import { ffmpegLogInitialOutput, safeKillFFmpeg } from "@scrypted/common/src/media-helpers";
import { readLength, readLine } from "@scrypted/common/src/read-stream";
import sdk, { FFmpegInput, Image, ImageOptions, MediaObject, ScryptedDeviceBase, ScryptedMimeTypes, VideoFrame, VideoFrameGenerator, VideoFrameGeneratorOptions } from "@scrypted/sdk";
import child_process from 'child_process';
import sharp from 'sharp';
import { Readable } from 'stream';
console.log(sharp.kernel);
sharp.kernel['linear'] = 'linear';
sharp.kernel['linear2'] = 'linear2';
async function createVipsMediaObject(image: VipsImage): Promise<VideoFrame & MediaObject> {
const ret = await sdk.mediaManager.createMediaObject(image, ScryptedMimeTypes.Image, {
timestamp: 0,
width: image.width,
height: image.height,
toBuffer: (options: ImageOptions) => image.toBuffer(options),
toImage: async (options: ImageOptions) => {
const newImage = await image.toVipsImage(options);
return createVipsMediaObject(newImage);
}
});
return ret;
}
interface RawFrame {
width: number;
height: number;
data: Buffer;
}
class VipsImage implements Image {
constructor(public image: sharp.Sharp, public width: number, public height: number) {
}
toImageInternal(options: ImageOptions) {
const transformed = this.image.clone();
if (options?.crop) {
transformed.extract({
left: Math.floor(options.crop.left),
top: Math.floor(options.crop.top),
width: Math.floor(options.crop.width),
height: Math.floor(options.crop.height),
});
}
if (options?.resize) {
transformed.resize(typeof options.resize.width === 'number' ? Math.floor(options.resize.width) : undefined, typeof options.resize.height === 'number' ? Math.floor(options.resize.height) : undefined, {
fit: "fill",
kernel: 'cubic',
});
}
return transformed;
}
async toBuffer(options: ImageOptions) {
const transformed = this.toImageInternal(options);
if (options?.format === 'rgb') {
transformed.removeAlpha().toFormat('raw');
}
else if (options?.format === 'jpg') {
transformed.toFormat('jpg');
}
return transformed.toBuffer();
}
async toVipsImage(options: ImageOptions) {
const transformed = this.toImageInternal(options);
const { info, data } = await transformed.raw().toBuffer({
resolveWithObject: true,
});
const newImage = sharp(data, {
raw: info,
});
const newMetadata = await newImage.metadata();
const newVipsImage = new VipsImage(newImage, newMetadata.width, newMetadata.height);
return newVipsImage;
}
async toImage(options: ImageOptions) {
if (options.format)
throw new Error('format can only be used with toBuffer');
const newVipsImage = await this.toVipsImage(options);
return createVipsMediaObject(newVipsImage);
}
}
export class FFmpegVideoFrameGenerator extends ScryptedDeviceBase implements VideoFrameGenerator {
async *generateVideoFramesInternal(mediaObject: MediaObject, options?: VideoFrameGeneratorOptions, filter?: (videoFrame: VideoFrame & MediaObject) => Promise<boolean>): AsyncGenerator<VideoFrame & MediaObject, any, unknown> {
const ffmpegInput = await sdk.mediaManager.convertMediaObjectToJSON<FFmpegInput>(mediaObject, ScryptedMimeTypes.FFmpegInput);
const args = [
'-hide_banner',
//'-hwaccel', 'auto',
...ffmpegInput.inputArguments,
'-vcodec', 'pam',
'-pix_fmt', 'rgb24',
'-f', 'image2pipe',
'pipe:3',
];
const cp = child_process.spawn(await sdk.mediaManager.getFFmpegPath(), args, {
stdio: ['pipe', 'pipe', 'pipe', 'pipe'],
});
ffmpegLogInitialOutput(this.console, cp);
let finished = false;
let frameDeferred: Deferred<RawFrame>;
const reader = async () => {
try {
const readable = cp.stdio[3] as Readable;
const headers = new Map<string, string>();
while (!finished) {
const line = await readLine(readable);
if (line !== 'ENDHDR') {
const [key, value] = line.split(' ');
headers[key] = value;
continue;
}
if (headers['TUPLTYPE'] !== 'RGB')
throw new Error(`Unexpected TUPLTYPE in PAM stream: ${headers['TUPLTYPE']}`);
const width = parseInt(headers['WIDTH']);
const height = parseInt(headers['HEIGHT']);
if (!width || !height)
throw new Error('Invalid dimensions in PAM stream');
const length = width * height * 3;
headers.clear();
const data = await readLength(readable, length);
if (frameDeferred) {
const f = frameDeferred;
frameDeferred = undefined;
f.resolve({
width,
height,
data,
});
}
else {
// this.console.warn('skipped frame');
}
}
}
catch (e) {
}
finally {
this.console.log('finished reader');
finished = true;
frameDeferred?.reject(new Error('frame generator finished'));
}
}
try {
reader();
while (!finished) {
frameDeferred = new Deferred();
const raw = await frameDeferred.promise;
const { width, height, data } = raw;
const image = sharp(data, {
raw: {
width,
height,
channels: 3,
}
});
const vipsImage = new VipsImage(image, width, height);
const mo = await createVipsMediaObject(vipsImage);
yield mo;
vipsImage.image.destroy();
vipsImage.image = undefined;
}
}
catch (e) {
}
finally {
this.console.log('finished generator');
finished = true;
safeKillFFmpeg(cp);
}
}
async generateVideoFrames(mediaObject: MediaObject, options?: VideoFrameGeneratorOptions, filter?: (videoFrame: VideoFrame & MediaObject) => Promise<boolean>): Promise<AsyncGenerator<VideoFrame & MediaObject, any, unknown>> {
return this.generateVideoFramesInternal(mediaObject, options, filter);
}
}

View File

@@ -1,10 +1,11 @@
import sdk, { Camera, DeviceState, EventListenerRegister, MediaObject, MediaStreamDestination, MixinDeviceBase, MixinProvider, MotionSensor, ObjectDetection, ObjectDetectionCallbacks, ObjectDetectionModel, ObjectDetectionResult, ObjectDetectionTypes, ObjectDetector, ObjectsDetected, ScryptedDevice, ScryptedDeviceType, ScryptedInterface, ScryptedMimeTypes, ScryptedNativeId, Setting, Settings, SettingValue, VideoCamera, VideoFrame, VideoFrameGenerator } from '@scrypted/sdk';
import sdk, { Camera, DeviceProvider, DeviceState, EventListenerRegister, MediaObject, MediaStreamDestination, MixinDeviceBase, MixinProvider, MotionSensor, ObjectDetection, ObjectDetectionCallbacks, ObjectDetectionModel, ObjectDetectionResult, ObjectDetectionTypes, ObjectDetector, ObjectsDetected, ScryptedDevice, ScryptedDeviceType, ScryptedInterface, ScryptedMimeTypes, ScryptedNativeId, Setting, Settings, SettingValue, VideoCamera, VideoFrame, VideoFrameGenerator } from '@scrypted/sdk';
import { StorageSettings } from '@scrypted/sdk/storage-settings';
import crypto from 'crypto';
import cloneDeep from 'lodash/cloneDeep';
import { AutoenableMixinProvider } from "../../../common/src/autoenable-mixin-provider";
import { SettingsMixinDeviceBase } from "../../../common/src/settings-mixin";
import { DenoisedDetectionEntry, DenoisedDetectionState, denoiseDetections } from './denoise';
import { FFmpegVideoFrameGenerator } from './ffmpeg-videoframes';
import { serverSupportsMixinEventMasking } from './server-version';
import { sleep } from './sleep';
import { getAllDevices, safeParseJson } from './util';
@@ -606,6 +607,7 @@ class ObjectDetectionMixin extends SettingsMixinDeviceBase<VideoCamera & Camera
this.console.error('video pipeline ended with error', e);
}
finally {
this.console.log('video pipeline analysis ended');
this.endObjectDetection();
}
}
@@ -1233,7 +1235,7 @@ class ObjectDetectorMixin extends MixinDeviceBase<ObjectDetection> implements Mi
}
}
class ObjectDetectionPlugin extends AutoenableMixinProvider implements Settings {
class ObjectDetectionPlugin extends AutoenableMixinProvider implements Settings, DeviceProvider {
currentMixins = new Set<ObjectDetectorMixin>();
storageSettings = new StorageSettings(this, {
@@ -1264,6 +1266,29 @@ class ObjectDetectionPlugin extends AutoenableMixinProvider implements Settings
constructor(nativeId?: ScryptedNativeId) {
super(nativeId);
process.nextTick(() => {
sdk.deviceManager.onDevicesChanged({
devices: [
{
name: 'FFmpeg Frame Generator',
type: ScryptedDeviceType.Builtin,
interfaces: [
ScryptedInterface.VideoFrameGenerator,
],
nativeId: 'ffmpeg',
}
]
})
})
}
async getDevice(nativeId: string): Promise<any> {
if (nativeId === 'ffmpeg')
return new FFmpegVideoFrameGenerator('ffmpeg');
}
async releaseDevice(id: string, nativeId: string): Promise<void> {
}
getSettings(): Promise<Setting[]> {