Add support for converting models with ImageProjectiveTransformV3 op

[maciej3031]

To see the logs from the Cloud Build CI, please join either our discussion or announcement mailing list.

This PR covers the following issue #6204

---

This change is 

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

For more information, open the CLA check for this pull request.

[lina128]

This is great, thank you for your contribution, @maciej3031 ! I only have some minor comments, then the PR should be ready to merge.

Reviewable status: 0 of 1 approvals obtained (waiting on @maciej3031)

---

tfjs-converter/python/tensorflowjs/op_list/image.json, line 114 at r1 (raw file):

      {
        "start": 0,
        "name": "image",

images, ref: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/ops/ops.pbtxt#L21603-L21650

Code quote:

image

---

tfjs-converter/python/tensorflowjs/op_list/image.json, line 124 at r1 (raw file):

      {
        "start": 2,
        "name": "outputShape",

Also add tfName: "output_shape", ref: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/ops/ops.pbtxt#L21603-L21650

Code quote:

outputShape

---

tfjs-converter/python/tensorflowjs/op_list/image.json, line 129 at r1 (raw file):

      {
        "start": 3,
        "name": "fillValue",

Also add tfName: "fill_value", ref: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/ops/ops.pbtxt#L21603-L21650

Code quote:

fillValue

---

tfjs-converter/src/operations/executors/image_executor.ts, line 82 at r1 (raw file):

        }
        case 'ImageProjectiveTransformV3': {
          const image =

images, it's a 4D, may take a batch of images.

Code quote:

image

[maciej3031]

Hi @lina128 , I addressed your comments except from these two:

- Also add tfName: "output_shape", ref: [...]
- Also add tfName: "fill_value", ref: [...]

I'm not very familiar with the tfjs repo but it seems to me that tfName does not exist in InputParamMapper so it cannot be used there. It's only present in AttrParamMapper 🤔

[lina128]

Ah, you are right.

[lina128]

Reviewable status: complete! 1 of 1 approvals obtained (waiting on @ahmedsabie and @maciej3031)