Return unbalanced latex equations

[GopinathCool]

I tested with lot of images, but for most of the images it results in an unbalanced latex equation, do you have any idea to resolve it?
I also attached the produced latex equation and image for reference.

$$U_{s2A_{-}k}=\bar{\cal B}{s}{s\bar{G}^{+}}\cdot\bar{Y}{s2A{-}\bar{k}}+\bar{\cal B}{r}\frac{\displaystyle\frac{\displaystyle\cal L}{m}^{2}}{\displaystyle\frac{\displaystyle\cal L}{\displaystyle\frac{\displaystyle\cal E}{m}}{\displaystyle\cal P}{s}\cdot\bar{\cal A}{2}\cdot\bar{\cal I}{s22}\cdot\bar{Y}{s2\bar{\cal B}{-}\bar{\cal F}{s}}{\displaystyle\cal E}{s2\bar{\cal F}{-}\displaystyle\cal A}^{-}\left(\frac{\displaystyle\frac{\displaystyle\frac{\displaystyle\cal E}{m}\cdot\bar{\cal S}{s}}{\displaystyle\cal E}{s}^{2}}\right)\cdot\bar{\mathrm Y}{s4\bar{\cal B}{s2}\cdot\bar{\cal A}{s2}\displaystyle\cal E{s}\bar{\cal F}_{s}\right},$$

[lukas-blecher]

Hi, thank you for your interest in the project.
I can see 2 or 3 problems with the Image:

1. The image is a bitmask (no gradient from black to white). The network has not trained on bitmask images.
2. Underscores are underrepresented in the training set
3. I don't recognize the font. Maybe the network has problems with that

When I rerender the image:

The model returns

U_{s t a k.\,k}=R_{s G2}\cdot Y_{s a A.\,k}+R_{r}\frac{L_{m}^{2}}{L_{r}^{2}}\cdot Y_{s1A F_{2}\,k}-\omega_{\psi s}\cdot\sigma_{2}L_{s2}\cdot Y_{s b P\,k}-\omega_{\psi s}\cdot\sigma_{2}L_{s a b F\,k}\cdot Y_{s1B F\,k}-\left(\frac{L_{m}\cdot R_{r}}{L_{r}^{2}}\right)\cdot\Psi_{r A k}

Which is not unbalanced and is fairly close to the input image.

I will take note and also train on a small subset of bitmask images. That should sort it out for the most part but will take some time.

[GopinathCool]

Thanks for your answer, the latex you have provided well balanced and also close to the image. But just one question how would be able to achieve that, do I need to update any part of the code?
It will be really helpful, if you could update and provide your model to handle bitmask images also.

[GopinathCool]

I have seen your recent commit, do I need to retrain the model with your formulae.zip images to handle bitmask images?

[lukas-blecher]

I have trained a new model (slightly different architecture) with the new commit 005648b.
The font is still hard to read for the model but it works. If you have problems try to change the temperature.