cospectrum

`model.run(inputs).unwrap()` failed ``` [.../onnx/src/ops/resize.rs:11] ctx.onnx_operator_set_version = 11 [.../onnx/src/ops/resize.rs:11] ctx.onnx_operator_set_version = 11 called `Result::unwrap()` on an `Err` value: Evaluating #219 "Resize_99" Resize: output 0, expected 1,24,40,40,F32, got 1,24,20,20,F32 ... ```

CI will pass after rebase I guess. But I think the number of images should be equal to the number of titles so that we don't do some weird magic....

Maybe we can just call it `filter`? 🤔 Although this is already inside a module called `filter`.

Or we can expose it to users as a Kernel method. `clamped_filter`, for example.

> Sure, either works. What would be the prefernce? Let's ask the owner @theotherphil. Do you have any preferences or maybe some criticism of this pr?

The strange thing is that this only works for odd sizes. I have an implementation of 2d convolution in Rust via "Im2Col" and matrix multiplication (compatible with [torch](https://pytorch.org/docs/stable/generated/torch.nn.functional.conv2d.html)), and the...

But what we can do about it? Find edge cases and `assert`?

I remember fixing the `text_size` function. It’s in the `master` branch. ```toml imageproc = { git = "https://github.com/image-rs/imageproc.git", branch = "master" } ```

> I remember fixing the `text_size` function. Try `master` branch. > > ```toml > imageproc = { git = "https://github.com/image-rs/imageproc.git", branch = "master" } > ``` Maybe it doesn’t solve...

For example, [TensorFlow](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/security/fuzzing/python_fuzzing.py#L169) python fuzzers use input bytes for shape only, values taken from RNG.