Skip to content

Commit

Permalink
use test_input instead of sample_input as default
Browse files Browse the repository at this point in the history
  • Loading branch information
@Nanguage
Nanguage committed Nov 14, 2023
1 parent e8da21c commit b643d3a
Showing 1 changed file with 94 additions and 50 deletions.
144 changes: 94 additions & 50 deletions src/components/TestRun.vue
View file
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,7 @@
<div id="buttons">
<b-button
:disabled="!this.buttonEnabledInput"
@click="this.loadSampleInput"
@click="this.loadTestInput"
>
Load sample image
</b-button>
Expand All @@ -26,7 +26,7 @@
</b-button>
<b-button
:disabled="!this.buttonEnabledOutput"
@click="this.loadSampleOutput"
@click="this.loadTestOutput"
>
Show reference output
</b-button>
Expand Down Expand Up @@ -60,7 +60,6 @@
#info-panel {
display: inline-block;
margin-left: 10px;
height: 10px;
margin-bottom: 20px;
}
.loader {
Expand All @@ -87,8 +86,9 @@
import { hyphaWebsocketClient } from "imjoy-rpc";
import * as tf from "@tensorflow/tfjs-core";
import "@tensorflow/tfjs-backend-cpu";
import npyjs from "npyjs";
function inferImgAxes(shape, order = "czb") {
function inferImgAxes(shape, order = "bcz") {
/**
* Infer the axes of an image.
*
Expand All @@ -109,12 +109,16 @@ function inferImgAxes(shape, order = "czb") {
}
}
function inferImgAxesViaSpec(shape, specAxes) {
let order = "czb";
if (!specAxes.includes("c")) {
order = "zb";
function inferImgAxesViaSpec(shape, specAxes, fromIJ = false) {
let order = "bcz";
if (fromIJ) {
order = "cz";
} else if (!specAxes.includes("c")) {
order = "bz";
} else if (!specAxes.includes("z")) {
order = "cb";
order = "bc";
} else if (!specAxes.includes("b")) {
order = "cz";
}
const imgAxes = inferImgAxes(shape, order);
return imgAxes;
Expand Down Expand Up @@ -178,7 +182,7 @@ Float64Array float64 float64
// return arrayBuffer;
//}
function toTfJs(arr) {
function ImjoyToTfJs(arr) {
if (arr._rvalue instanceof ArrayBuffer) {
arr._rvalue = new Uint8Array(arr._rvalue);
}
Expand All @@ -195,12 +199,6 @@ function toTfJs(arr) {
}
}
}
//const Constructor = getConstructor(arr._rdtype);
//const bLen = multiplyArrayElements(arr._rshape) * Constructor.BYTES_PER_ELEMENT;
//const bytes = arr._rvalue.buffer.slice(0, bLen)
////const reversedBytes = reverseEndianness(bytes, Constructor.BYTES_PER_ELEMENT);
////const tarr = new Constructor(reversedBytes);
//const tarr = new Constructor(bytes)
const tensor = tf.tensor(Array.from(tarr), arr._rshape);
tensor._rdtype = arr._rdtype;
return tensor;
Expand Down Expand Up @@ -390,14 +388,20 @@ function processForShow(img, specAxes) {
[z-stack, height, width, 1]
[z-stack, height, width, 3] (will show as a stack of RGB image)
*/
const tensor = toTfJs(img);
const tensor = ImjoyToTfJs(img);
const splitedArrs = splitForShow(tensor, specAxes);
return splitedArrs.map(arr => {
arr._rdtype = tensor._rdtype;
return toImJoyArr(arr);
});
}
function rdfHas(rdf, key) {
return rdf[key] !== undefined && rdf[key].length > 0;
}
export default {
name: "TestRunForm",
props: {
Expand Down Expand Up @@ -429,16 +433,10 @@ export default {
await this.loadImageJ();
this.setInfoPanel("");
this.buttonEnabledRun = true;
const hasSampleInput =
this.rdf.sample_inputs !== undefined &&
this.rdf.sample_inputs.length > 0;
if (hasSampleInput) {
if (rdfHas(this.rdf, "test_inputs") || rdfHas(this.rdf, "sample_inputs")) {
this.buttonEnabledInput = true;
}
const hasSampleOutput =
this.rdf.sample_outputs !== undefined &&
this.rdf.sample_outputs.length > 0;
if (hasSampleOutput) {
if (rdfHas(this.rdf, "test_outputs") || rdfHas(this.rdf, "sample_outputs")) {
this.buttonEnabledOutput = true;
}
},
Expand Down Expand Up @@ -481,10 +479,10 @@ export default {
const inputSpec = this.rdf.inputs[0];
await this.api.log("Spec input axes: " + inputSpec.axes);
let imgAxes;
imgAxes = inferImgAxesViaSpec(img._rshape, inputSpec.axes);
imgAxes = inferImgAxesViaSpec(img._rshape, inputSpec.axes, true);
await this.api.log("Input image axes: " + imgAxes);
await this.api.log("Reshape image to match the input spec.");
const tensor = toTfJs(img);
const tensor = ImjoyToTfJs(img);
const newTensor = mapAxes(tensor, imgAxes, inputSpec.axes);
const reshapedImg = toImJoyArr(newTensor);
const resp = await this.bioengineExecute(this.resourceItem.id, [
Expand All @@ -498,14 +496,19 @@ export default {
debugger;
return;
}
this.setInfoPanel("Success!");
this.setInfoPanel("");
const outImg = resp.result.outputs[0];
await this.api.log("Output image shape: " + outImg._rshape);
const outputSpec = this.rdf.outputs[0];
await this.api.log("Spec output axes: " + outputSpec.axes);
const imgsForShow = processForShow(outImg, outputSpec.axes);
for (let i = 0; i < imgsForShow.length; i++) {
const img = imgsForShow[i];
await this.showImgs(imgsForShow);
this.buttonEnabledRun = true;
},
async showImgs(imgs) {
for (let i = 0; i < imgs.length; i++) {
const img = imgs[i];
await this.api.log(
"Output image shape after processing: " + img._rshape
);
Expand All @@ -517,7 +520,6 @@ export default {
}
// await this.ij.runMacro("run('Enhance Contrast', 'saturated=0.35');")
}
this.buttonEnabledRun = true;
},
async loadRdf() {
Expand Down Expand Up @@ -557,36 +559,78 @@ export default {
async viewImgFromUrl(url) {
await this.api.log("View image from url: " + url);
const resp = await fetch(url);
if (!resp.ok) {
this.setInfoPanel("Failed to load the image.");
console.error(resp);
return;
}
const arrayBuffer = await resp.arrayBuffer();
let fileName;
if (url.endsWith("/content")) {
fileName = url.split("/")[url.split("/").length - 2];
} else {
fileName = url.split("/")[url.split("/").length - 1];
}
this.ij.viewImage(arrayBuffer, { name: fileName }).catch(err => {
console.error(err);
this.setInfoPanel("Failed to view the image.");
});
if (fileName.endsWith(".npy")) {
let nj = new npyjs();
const res = await nj.load(url);
const value = new Uint8Array(
res.data.buffer.slice(res.data.byteOffset));
const imjArr = {
_rtype: "ndarray",
_rdtype: res.dtype,
_rshape: res.shape,
_rvalue: value,
};
const inputSpec = this.rdf.inputs[0];
const imgAxes = inferImgAxesViaSpec(
imjArr._rshape, inputSpec.axes);
const imgsForShow = processForShow(imjArr, imgAxes);
await this.showImgs(imgsForShow)
} else {
const resp = await fetch(url);
if (!resp.ok) {
this.setInfoPanel("Failed to load the image.");
console.error(resp);
return;
}
const arrayBuffer = await resp.arrayBuffer();
this.ij.viewImage(arrayBuffer, { name: fileName }).catch(err => {
console.error(err);
this.setInfoPanel("Failed to view the image.");
});
}
},
async loadSampleInput() {
const url = this.rdf.sample_inputs[0];
this.setInfoPanel("Loading sample input...", true);
await this.viewImgFromUrl(url);
async loadTestInput() {
this.setInfoPanel("Loading test input...", true);
if (rdfHas(this.rdf, "test_inputs")) {
try {
await this.viewImgFromUrl(this.rdf.test_inputs[0]);
} catch (err) {
await this.api.log("Failed to load the test input.");
console.error(err);
await this.api.log("Loading sample input instead...");
await this.viewImgFromUrl(this.rdf.sample_inputs[0]);
}
} else if (rdfHas(this.rdf, "sample_inputs")) {
await this.viewImgFromUrl(this.rdf.sample_inputs[0]);
} else {
await this.api.alert("No test input found.");
}
this.setInfoPanel("");
},
async loadSampleOutput() {
const url = this.rdf.sample_outputs[0];
this.setInfoPanel("Loading sample output...", true);
await this.viewImgFromUrl(url);
async loadTestOutput() {
this.setInfoPanel("Loading test output...", true);
if (rdfHas(this.rdf, "test_outputs")) {
try {
await this.viewImgFromUrl(this.rdf.test_outputs[0]);
} catch (err) {
await this.api.log("Failed to load the test output.");
console.error(err);
await this.api.log("Loading sample output instead...");
await this.viewImgFromUrl(this.rdf.sample_outputs[0]);
}
} else if (rdfHas(this.rdf, "sample_outputs")) {
await this.viewImgFromUrl(this.rdf.sample_outputs[0]);
} else {
await this.api.alert("No test output found.");
}
this.setInfoPanel("");
}
}
Expand Down

0 comments on commit b643d3a

Please sign in to comment.