search

zldrobit / tfjs-yolov5-example

TensorFlow.js example for YOLOv5
https://codesandbox.io/s/github/zldrobit/tfjs-yolov5-example
89 stars 30 forks source link

Unhandled Rejection (TypeError): valid_detections is undefined #3

Open Danielvenzi opened 3 years ago

Danielvenzi commented 3 years ago

Hello, I'm trying to test the yolov5s to tensorflow.js conversion with a example image of the default 320x320 size, but when i put it on the "Choose or drop a file" at the web app i receive immediately the "valid_detections is undefined" error on the index.js:82 .

79 | 80 | const [boxes, scores, classes, valid_detections] = res; 81 | var i; -> 82 | for (i = 0; i < valid_detections.dataSync()[0]; ++i){ | ^ let [x1, y1, x2, y2] = boxes.dataSync().slice(i 4, (i + 1) 4); 84 | x1 = c.width; 85 | x2 = c.width;

I'm using the default yolov5s model to test this functionality with a image that has classes that the model can detect, i resized the image previous to the test in order to avoid errors, but i don't seem to know why this problem is happening. Do you have any idea?

zldrobit commented 3 years ago

Hello. valid_detections is undefined means that this variable is not assigned. You could add a breakpoint on the line

80 | const [boxes, scores, classes, valid_detections] = res;

, and inspect the values of boxes, scores, classes, valid_detections to check whether they are assigned.

zldrobit commented 3 years ago

@Danielvenzi This issue may be related to https://github.com/zldrobit/tfjs-yolov5-example/issues/2. Plz update tfjs to 3.9.0 and try again. You could use the latest commit of this repo and run npm install to upgrade tfjs.