Aleksei Zinovev

Add Tile function

2

Hi, I'm going to implement the postprocessing for the YOLOv4 model in the KotlinDL library, but it requires many different analogs of the NumPy functions. Could you please add the...

Make old versions from jcenter are available on MavenCentral

1

Due to jcenter closing from, old version like 5.0.1 will be not available in the nearest future. We are using klaxon in [KotlinDL project](https://github.com/JetBrains/KotlinDL). After migration to 5.4, we've got...

- [ ] with alternative way to load the model from GDrive or Github large files like in the ONNX/models - [ ] with the short description of how the...

There are a few noTop models available in the ONNX Model Hub and all of them could be fine-tuned with the TensorFlow added layers like Dense or AveragePool2D. There is...

The existing toString method just prints some underlying information from the ONNX Java API. We need to unify the input/output information with the TF module and also add info from...

The initializer name schema is violated by Zero and Ones ops

1

If you will load a graph from Keras in hdf5 format it could keep Zeros or Ones as a default initializer for biases. But the standard schema of loading weights...

Replace the BufferedImage on Multik NDarrays

2

This is a research tickets, the MVP solution should include the following things: - [ ] Merge transformImage/transformTensor DSL parts - [ ] Remove BufferedImage from signatures - [ ]...

Add input/outputShape information to all KDocs of layers

2

We need to have better documentation for our layers. A few layers have inputShape/outputShape information in the KDocs, copied from Keras documentation, for example [GlobalAvgPool2D](https://github.com/JetBrains/KotlinDL/blob/41f0b20d3bfd6e277d8486d945cdef59a6250b9b/api/src/main/kotlin/org/jetbrains/kotlinx/dl/api/core/layer/pooling/GlobalAvgPool2D.kt#L20) Please add the same documentation...

Add EfficientNetV2 model family to the ONNX Model Hub

1

Add to the ONNXModelHub the [7 new models](https://www.tensorflow.org/api_docs/python/tf/keras/applications/efficientnet_v2), each model should be ported to the ONNX format with tf2onnx project in two variants (top and no_top)