Initial codebase (#1)

* Add project code * Logger improvements * Improvements to web demo code * added create_wlasl_landmarks_dataset.py and xtract_mediapipe_landmarks.py * Fix rotation augmentation * fixed error in docstring, and removed unnecessary replace -1 -> 0 * Readme updates * Share base notebooks * Add notebooks and unify for different datasets * requirements update * fixes * Make evaluate more deterministic * Allow training with clearml * refactor preprocessing and apply linter * Minor fixes * Minor notebook tweaks * Readme updates * Fix PR comments * Remove unneeded code * Add banner to Readme --------- Co-authored-by: Gabriel Lema <gabriel.lema@xmartlabs.com>
2023-03-03 10:07:54 -03:00
parent 661e4bbc03
commit 81bbf66aab
49 changed files with 4205 additions and 0 deletions
--- a/web/README.md
+++ b/web/README.md
@@ -0,0 +1,8 @@
+# SPOTER Web
+
+To test Spoter model in the web, follow these steps:
+* Convert your latest Pytorch model to Onnx by running `python convert.py`. This is best done inside the Docker container. You will need to install additional dependencies for the conversions (see commented lines in requirements.txt)
+* The ONNX should be generated in the `web` folder, otherwise copy it there.
+* run `npx light-server -s . -p 8080` in the `web` folder. (`npx` comes with `npm`)
+
+Enjoy!
--- a/web/index.html
+++ b/web/index.html
@@ -0,0 +1,61 @@
+<!DOCTYPE html>
+<html>
+    <header>
+        <title>ONNX Runtime JavaScript examples: Quick Start - Web (using script tag)</title>
+    </header>
+    <body>
+        <button id="start-test">Start Test</button>
+        <p id="output"></p>
+        <!-- import ONNXRuntime Web from CDN -->
+        <script src="https://cdn.jsdelivr.net/npm/onnxruntime-web/dist/ort.min.js"></script>
+        <script>
+            async function setupButtons() {
+                let test_button = document.querySelector("#start-test");
+
+                test_button.addEventListener('click', async function() {
+                    main();
+                });
+            }
+            // use an async context to call onnxruntime functions.
+            async function main() {
+                try {
+                    // create a new session and load the specific model.
+                    //
+                    // the model in this example contains a single MatMul node
+                    // it has 2 inputs: 'a'(float32, 3x4) and 'b'(float32, 4x3)
+                    // it has 1 output: 'c'(float32, 3x3)
+                    const session = await ort.InferenceSession.create('./spoter.onnx');
+
+                    // Number of frames
+                    const N = 100
+
+                    // prepare inputs. a tensor need its corresponding TypedArray as data
+                    const dataA = new Float32Array(108 * N);
+                    dataA.fill(0.4);
+                    // const dataB = Float32Array.from([10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120]);
+                    const tensorA = new ort.Tensor('float32', dataA, [1, N, 54, 2]);
+                    console.log(tensorA);
+
+                    // prepare feeds. use model input names as keys.
+                    const feeds = { input: tensorA };
+
+                    // feed inputs and run
+                    startTime = new Date();
+                    const results = await session.run(feeds);
+                    // read from results
+                    const dataC = results.output.data;
+                    endTime = new Date();
+                    let output = document.querySelector("#output");
+
+                    var timeDiff = endTime - startTime; //in ms
+                    output.innerText = `Data of result tensor 'output':\n ${dataC}` + "\nInference took " + timeDiff + " ms";
+
+                } catch (e) {
+                    let output = document.querySelector("#output");
+                    output.innerText = `failed to inference ONNX model: ${e}.`;
+                }
+            }
+            setupButtons();
+        </script>
+    </body>
+</html>