Can't use in a lerna/create-react app project #5
Comments
|
What are your Looks like I've compiled Face-API using newer ECMA standard that your project does not recognize. Original Face-API used ES2015 and I'm using ES2018. If you're using TypeScript, you can try importing TS sources directly so it gets rebuild along your application: Or check docs on how to rebuild Face-API after modifying it's |
|
Thanks for the quick reply! Can I just change my tsconfig to use ES2018 instead? |
|
Sure 😁 |
|
using es2018 I get a better error message now :) ../face-detect-plugin/node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Abs.js Questions:
Thanks |
|
Actually the console have many more errors regarding not found modules: log.js:24 [HMR] Waiting for update signal from WDS... |
|
Q: Why can't it just work like the original face-api? TFJS 2.x is not fully compatible with TFJS 1.x. Q: Why need to install both face-api and tfjs if tfjs is a dependency of face-api? TFJS 1.x was a simple package. Q: Why with the original face-api I don't need to import tfjs for it to compile? In short, I though that bundling entire TFJS 2.x makes But as you've stated (and others as well), people prefer simplicity over modularity so I'll make it so. I'll publish a new version of FaceAPI shortly that includes TFJS and does not require any extra imports. |
|
Hi, I really appreciate your quick and detailed responses. I actually like tfjs not bundled since I am using it in another audio related detection code and prefer to control which version I use. I was just wondering since I saw it is a dependency in your package.json and not a peer dependency. Do you have any idea how to resolve the errors? It seems the code cannot locate the core and other tf libraries. I can wait for your new version if its coming soon or find a workaround like including it in the html as a script so the bundler will not process it? |
|
An another question, is there a chance of using face-api with webworkers so it won't block the UI thread? I know its possible with tfjs itself. |
|
New version is almost ready, I'm updating the docs now and should be published within an hour. Regarding WebWorkers - It's possible, but somewhat outside of the scope of what I've intended with this port of FaceAPI (it's already much more work than I envisioned). One of the issues is that FaceAPI internally uses standard I'll try to include that in my new library (hopefully by the end of this week). |
|
thanks, that will be amazing! |
|
New version |
|
Cannot find module '@vladmandic/face-api' or its corresponding type declarations. |
|
silly me, that's what happens when you rush and publish without proper testing. republished - 0.7.2. |
|
same I think because "typings": "build/index.d.ts" is actually in src? |
|
it's in |
|
Now I don't need to import tfjs before face-api but I get this error now because the faceapi returned from the import is an empty object {}. Also I think now tfjs should be a dependency and not devDependeny or peerDepenedency. Also you should probably use tfsj as npm module and use the dist/tf.es2017.js. Probably need to take a step back and rethink the whole module thing. Most people these days use modules and a. bundler. TypeError: faceapi.TinyFaceDetectorOptions is not a constructor
|
|
Ok could make it work with import * as faceapi from '@vladmandic/face-api/dist/face-api.esm' and then declare module "@vladmandic/face-api/dist/face-api.esm" But lost all the typings, it is ok for now but we will happy to help get the module stuff to work with types Thanks! |
|
You were correct:
{
// old default
"browser": "dist/face-api.js",
// should be
"browser": "dist/face-api.esm.js",
}and as you saw, you can manually specify which version to include: // will import default
import * as faceapi from '@vladmandic/face-api';
// esm with tfjs bundled, this is now the default
import * as faceapi from '@vladmandic/face-api/dist/face-api.esm.js';
// esm without tfjs bundled, you must manually include @tensorflow/tfjs before importing face-api
import * as faceapi from '@vladmandic/face-api/dist/face-api.nobundle.js';as you can see, I've now added both prebundled and nobundle versions, you can pick which one you want to use. regarding dependencies, it only means should |
|
Will I get the typings this way? |
|
it should. |
|
it works now! never had such responsiveness in an open source project :) thanks What is the new library, maybe face-api with webworkers? |
|
thanks :) new library is using totally different models and is written from scratch - i found that less frustrating than trying to modify you can take a look at a prerelease: https://github.com/vladmandic/human |
|
closing as resolved |
I am using the library in a TypeScript lerna module which is build with tsc and then bundled as part of a create-react-app. I had no issue with the original face-api but wanted the latest fdjs. I am using the npm module with npm install.
If I only do:
import * as faceapi from '@vladmandic/face-api'
I get a compile error:
Failed to compile
../face-detect-plugin/node_modules/@vladmandic/face-api/dist/face-api.js 344:47
Module parse failed: Unexpected token (344:47)
File was processed with these loaders:
You may need an additional loader to handle the result of these loaders.
| } : T || {};
|
| F = new d(_),
| D = function () {
if I add:
import * as tf from '@tensorflow/tfjs'
before importing face-api then I get an error in the console:
Uncaught SyntaxError: Unexpected token '!'
If I look at the code I see this:
class MathBackendCPU extends !(function webpackMissingModule() { var e = new Error("Cannot find module '@tensorflow/tfjs-core'"); e.code = 'MODULE_NOT_FOUND'; throw e; }()) {
constructor() {
super();
this.blockSize = 48;
this.firstUse = true;
this.data = new !(function webpackMissingModule() { var e = new Error("Cannot find module '@tensorflow/tfjs-core'"); e.code = 'MODULE_NOT_FOUND'; throw e; }())(this, !(function webpackMissingModule() { var e = new Error("Cannot find module '@tensorflow/tfjs-core'"); e.code = 'MODULE_NOT_FOUND'; throw e; }())());
}
The text was updated successfully, but these errors were encountered: