Deep Learning Blog

Trackformer — Multi-Object Tracking with Transformers

A lower-level explanation of the paper Multi-Object Tracking with Transformers (Meinhardt et al., 2022) including DETR (Carion et al., 2020) and Deformable DETR (Zhu et al., 2020). 🎥

25 min read   ·   March 8, 2023

2023   ·   applications   transformer   MOT   multiple-object-tracking  

8. Transformer

⏮️ I recommend reading the RNN post first for the encoder-decoder architecture.

18 min read   ·   March 8, 2023

2023   ·   dl-fundamentals   neural-net-archetype   attention   transformer   generative-models  

7. Recurrent Neural Networks

🧱 The encoder-decoder RNN is quite useful for understanding the transformer.

9 min read   ·   March 7, 2023

2023   ·   dl-fundamentals   sneural-net-archetype   backprop   exploding-vanishing-gradients  

6. Convolutional Neural Networks

🔑️ Key sections about parameter sharing, inductive biases, skip connections, and cross-correlation.

15 min read   ·   March 6, 2023

2023   ·   dl-fundamentals   neural-net-archetype   backprop   cross-correlation   inductive-bias   receptive-field   compute-graph  

5. Feedforward Neural Networks

6 min read   ·   March 5, 2023

2023   ·   dl-fundamentals   neural-net-archetype   backprop  

4. Backpropagation

💡 Includes a nice detail about why we don't 'frontpropagate'. ;-)

10 min read   ·   March 4, 2023

2023   ·   dl-fundamentals   backprop   reverse-accumulation   compute-graph  

3. Optimization

6 min read   ·   March 3, 2023

2023   ·   dl-fundamentals   optimization   first-order-methods  

2. Supervised Learning

11 min read   ·   March 2, 2023

2023   ·   dl-fundamentals   supervised-learning