feat(post): add alquimodelia post

_config.yml

@@ -36,6 +36,14 @@ authors:
     web: https://github.com/vitornvpaixao
     twitter: https://github.com/vitornvpaixao
     description: "Frontend programmer and forever learner. Has been in charge of maintaining and extending our Vue.Js app."
+  joao:
+    name: João Santos
+    display_name: João
+    avatar: 'assets/images/joao.jpeg'
+    web: https://github.com/JotaFan
+    twitter: https://github.com/JotaFan
+    description: "Versatile programmer, deep learning modeling wizzard, and geospatial analyist. Has contributed to the pixels platform and executed many projects for Tesselo."
+
 
 # Plugins
 plugins:

_posts/2023-02-28-our-models.md

@@ -0,0 +1,70 @@
+---
+layout: post
+title:  "Model Alquemy"
+author: joao
+categories: [ AI, code ]
+image: assets/images/our-models.png
+description: "An introduction to Tesselo's AI modeling, explaining the model types we used for our mapping with EO data."
+featured: false
+hidden: false
+---
+Tesselo's most successful deep learning models are presented in this post. We have
+used them to do large scale land cover modeling across the world.
+
+We have packaged our most common models into a repository that makes it easy
+to use them. You can find the model references in our
+[Alquimodelia](https://github.com/tesselo/alquimodelia) repository. It contains
+the detailed model definitions for our most successful models. We used Keras with a
+Tensorflow backend for our modeling, so the definitions are written in that famework.
+
+The aim of Alquimodelia is to provide a user friendly way to use and change parameters
+on the common model architectures used in Tesselo. The model classes can be created without
+deep knowledge of keras or tensorflow. The main required parameters are the input and ouput
+shape that the models will work with. Then, Arquimodelia will construct the models accordingly.
+
+## Model types
+
+Depending on the context and the goal of the modeling, we have used a series of
+different models. They range from pixel based classifiers to time-series based
+U-Net type architectures.
+
+## Use all bands
+
+The multispectral imagery allows us to get context from the various ranges of light available. This could help overcome coud coverage over a site, or could easily identify Land Uses.
+The usage of multiple bands allowed our models to freely aquire the best out of Open Data targeting the goal at hands.
+
+## Classifiers
+
+Here we are giving a quick overview of the different model types and their use cases.
+Detailed posts about some of the models will follow separately as well.
+
+### Pixel based time series classifier
+
+This classifier is quite small but very powerful for small training datasets. It is
+non-sequential and based on one-dimensional convolution. It has two branches that
+are detecting patterns in time series at different levels.
+
+### Single scene image segmentation
+
+2D U-Net or ResNet based.
+
+#### ResNet
+
+The ResNet architecture uses two-dimensional convolutions to provide a classification to
+a given image. This has been used as a way to classify images with a single class. Or it
+could be used to classify a single pixel, but with the context of the surroundings.
+
+#### 2D U-Net
+
+The 2D U-Net is similiar to ResNet in terms of the usage of two-dimensional convolutions, but
+instead of giving one answer for each image, it responds with also an image.
+Used in image classification and segmentation.
+
+### Time series of images
+
+#### 3D U-Net
+
+The 3D U-Net architeture follows the same patterns and the 2D, but instead of two-dimensional
+convolutions it uses three-dimensional convolutions, multiple images across time.
+The answer would still be a single image, but produced with time context.
+Great to surpass problems like clouds and other imagery artifacts.