**NeuralRecon** is a framework for real-time 3D scene reconstruction from a monocular video. Unlike previous methods that estimate single-view depth maps separately on each key-frame and fuse them later, NeuralRecon proposes to directly reconstruct local surfaces represented as sparse TSDF volumes for each video fragment sequentially by a neural network. A learning-based TSDF fusion module based on gated recurrent units is used to guide the network to fuse features from previous fragments. This design allows the network to capture local smoothness prior and global shape prior of 3D surfaces.

**Linear Warmup With Linear Decay** is a learning rate schedule in which we increase the learning rate linearly for $n$ updates and then linearly decay afterwards.

Linear Warmup With Linear Decay

NeuralRecon

NeuralRecon: Real-Time Coherent 3D Reconstruction from Monocular Video

**Entropy Regularization** is a type of regularization used in [reinforcement learning](https://paperswithcode.com/methods/area/reinforcement-learning). For on-policy policy gradient based methods like [A3C](https://paperswithcode.com/method/a3c), the same mutual  reinforcement behaviour leads to a highly-peaked $\pi\left(a\mid{s}\right)$ towards a few actions or action sequences, since it is easier for the actor and critic to overoptimise to a small portion of the environment. To reduce this problem, entropy regularization adds an entropy term to the loss to promote action diversity:

$$H(X) = -\sum\pi\left(x\right)\log\left(\pi\left(x\right)\right) $$

Image Credit: Wikipedia

Source	NeuralRecon: Real-Time Coherent 3D Reconstruction from Monocular Video
Year	2000
Data Source	CC BY-SA - https://paperswithcode.com