Node.js support

[mycaule]

Small manual verifications :

• Ability to import the package directly from NPM
• regressionCheck.ts still runs fine and is representative of the NPM use case
• Code samples with express (popular framework) and sharp (different image file formats)

Overall the code was already fine, I just added information and helper functions in README.md since converting to Tensor3d and convert file formats can be hard for someone unfamiliar.

I only wasn't able to nsfw.load(...) the TensorflowJS 299x299 Image Model files, they seem different from what is loaded by default TensorflowJS Quantized 224x224 Image Model.

Since it is interesting to load bigger models in server side comparing to client side, you might also want to provide information on how to load the three other bigger models at:
https://github.com/GantMan/nsfw_model

• ❌TensorflowJS 299x299 Image Model
• TensorflowJS Quantized 299x299 Image Model
• TensorflowJS 224x224 Image Model
• ✔️TensorflowJS Quantized 224x224 Image Model

Ideally these 4 should be tested with Jest too, I don't know if you plan to maintain them or update them depending on how the scrapper repo updates.

[GantMan]

Looks great! Will merge. I might make some helper code built into the module for servers later.

[mycaule]

The only problem with changing the model is that there is a magic constant 224 vs 299 somewhere I guess. It might be easy to change in the code.

expected input_1 to have shape [null,299,299,3] but got array with shape [1,224,224,3].

require('nsfwjs').load('file://./path/to/model/', {size: 299}) solved the problem.

[GantMan]

👍 - btw, just added you to the README as a contributor.