# sanvisharma850/intentformer-trajectory-prediction architecture

> Architecture diagram of the GitHub repository sanvisharma850/intentformer-trajectory-prediction, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-07). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/sanvisharma850/intentformer-trajectory-prediction
- Repository: https://github.com/sanvisharma850/intentformer-trajectory-prediction

## Overview

IntentFormer is a notebook-based research workflow for predicting pedestrian and cyclist trajectories from nuScenes. The documented path runs from dataset preprocessing to model training and evaluation; the model combines temporal, social, and LiDAR context to produce multimodal trajectories and intent labels. The graph preserves the README’s documented model components, losses, and metrics. Source excerpts were not available, so notebook internals and some handoffs are not verified; those links are dashed where inferred.

Main entry points: [`02-model-train-v2.ipynb`](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb) (Model training), [`01-dataprep-v2.ipynb`](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/01-dataprep-v2.ipynb) (Data preprocessing), [`03-eval-v2.ipynb`](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/03-eval-v2.ipynb) (Model evaluation), [`Training_Output/evaluation_summary.md`](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/Training_Output/evaluation_summary.md) (Evaluation results)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_prep["Data preparation"]
  node_dataprep["Data preprocessing"]
  node_trajdata["Trajectories"]
  node_neighbor_data["Neighbor data"]
  node_lidar_data["LiDAR features"]
  node_norm_stats["Normalization stats"]
end

subgraph group_model["Prediction model"]
  node_temporal["Temporal encoder"]
  node_social["Social encoder"]
  node_lidar_encoder["LiDAR features"]
  node_trajectory_head["Trajectory modes"]
  node_intent_head["Intent classifier"]
  node_gmm["GMM predictions"]
end

subgraph group_training["Training"]
  node_training["Model training"]
  node_gmm_loss["GMM-NLL loss"]
  node_diversity_loss["Diversity loss"]
  node_intent_loss["Intent CE loss"]
end

subgraph group_evaluation["Evaluation"]
  node_evaluation["Model evaluation<br/>[03-eval-v2.ipynb]"]
  node_metrics["Prediction metrics<br/>[03-eval-v2.ipynb]"]
  node_results["Evaluation results"]
end

node_dataset[("nuScenes dataset")]

node_dataset -->|"provides data"| node_dataprep
node_dataprep -->|"produces"| node_trajdata
node_dataprep -->|"produces"| node_neighbor_data
node_dataprep -->|"produces"| node_lidar_data
node_dataprep -->|"produces"| node_norm_stats
node_temporal -->|"encodes history"| node_trajectory_head
node_social -->|"provides context"| node_trajectory_head
node_lidar_encoder -->|"provides features"| node_trajectory_head
node_trajectory_head -->|"predicts modes"| node_gmm
node_intent_head -->|"classifies intent"| node_gmm
node_training -->|"optimizes with"| node_gmm_loss
node_training -->|"optimizes with"| node_diversity_loss
node_training -->|"optimizes with"| node_intent_loss
node_training -->|"trains model"| node_temporal
node_training -->|"trains model"| node_social
node_training -->|"trains model"| node_lidar_encoder
node_training -->|"trains model"| node_trajectory_head
node_training -->|"trains model"| node_intent_head
node_trajdata -.->|"feeds training"| node_training
node_neighbor_data -.->|"feeds training"| node_training
node_lidar_data -.->|"feeds training"| node_training
node_norm_stats -.->|"feeds training"| node_training
node_evaluation -->|"computes"| node_metrics
node_evaluation -.->|"records results"| node_results
node_gmm -.->|"evaluates predictions"| node_evaluation

click node_dataprep "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/01-dataprep-v2.ipynb"
click node_trajdata "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/01-dataprep-v2.ipynb"
click node_neighbor_data "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/01-dataprep-v2.ipynb"
click node_lidar_data "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/01-dataprep-v2.ipynb"
click node_norm_stats "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/01-dataprep-v2.ipynb"
click node_temporal "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb"
click node_social "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb"
click node_lidar_encoder "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb"
click node_trajectory_head "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb"
click node_intent_head "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb"
click node_gmm "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb"
click node_training "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb"
click node_gmm_loss "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb"
click node_diversity_loss "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb"
click node_intent_loss "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb"
click node_evaluation "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/03-eval-v2.ipynb"
click node_metrics "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/03-eval-v2.ipynb"
click node_results "https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/Training_Output/evaluation_summary.md"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_dataprep,node_trajdata,node_neighbor_data,node_lidar_data,node_norm_stats toneBlue
class node_temporal,node_social,node_lidar_encoder,node_trajectory_head,node_intent_head,node_gmm,node_dataset toneAmber
class node_training,node_gmm_loss,node_diversity_loss,node_intent_loss toneMint
class node_evaluation,node_metrics,node_results toneRose
```

## Components

### Data preparation

- [Data preprocessing](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/01-dataprep-v2.ipynb): `01-dataprep-v2.ipynb`
- [Trajectories](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/01-dataprep-v2.ipynb): `01-dataprep-v2.ipynb`
- [Neighbor data](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/01-dataprep-v2.ipynb): `01-dataprep-v2.ipynb`
- [LiDAR features](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/01-dataprep-v2.ipynb): `01-dataprep-v2.ipynb`
- [Normalization stats](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/01-dataprep-v2.ipynb): `01-dataprep-v2.ipynb`

### Prediction model

- [Temporal encoder](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb): `02-model-train-v2.ipynb`
- [Social encoder](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb): `02-model-train-v2.ipynb`
- [LiDAR features](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb): `02-model-train-v2.ipynb`
- [Trajectory modes](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb): `02-model-train-v2.ipynb`
- [Intent classifier](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb): `02-model-train-v2.ipynb`
- [GMM predictions](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb): `02-model-train-v2.ipynb`

### Training

- [Model training](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb): `02-model-train-v2.ipynb`
- [GMM-NLL loss](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb): `02-model-train-v2.ipynb`
- [Diversity loss](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb): `02-model-train-v2.ipynb`
- [Intent CE loss](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/02-model-train-v2.ipynb): `02-model-train-v2.ipynb`

### Evaluation

- [Model evaluation](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/03-eval-v2.ipynb): `03-eval-v2.ipynb`
- [Prediction metrics](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/03-eval-v2.ipynb): `03-eval-v2.ipynb`
- [Evaluation results](https://github.com/sanvisharma850/intentformer-trajectory-prediction/blob/main/Training_Output/evaluation_summary.md): `Training_Output/evaluation_summary.md`

### Other components

- **nuScenes dataset**

## Connections

- nuScenes dataset → Data preprocessing: provides data (evidence: `README.md`)
- Data preprocessing → Trajectories: produces (evidence: `README.md`)
- Data preprocessing → Neighbor data: produces (evidence: `README.md`)
- Data preprocessing → LiDAR features: produces (evidence: `README.md`)
- Data preprocessing → Normalization stats: produces (evidence: `README.md`)
- Temporal encoder → Trajectory modes: encodes history (evidence: `README.md`)
- Social encoder → Trajectory modes: provides context (evidence: `README.md`)
- LiDAR features → Trajectory modes: provides features (evidence: `README.md`)
- Trajectory modes → GMM predictions: predicts modes (evidence: `README.md`)
- Intent classifier → GMM predictions: classifies intent (evidence: `README.md`)
- Model training → GMM-NLL loss: optimizes with (evidence: `README.md`)
- Model training → Diversity loss: optimizes with (evidence: `README.md`)
- Model training → Intent CE loss: optimizes with (evidence: `README.md`)
- Model training → Temporal encoder: trains model (evidence: `README.md`)
- Model training → Social encoder: trains model (evidence: `README.md`)
- Model training → LiDAR features: trains model (evidence: `README.md`)
- Model training → Trajectory modes: trains model (evidence: `README.md`)
- Model training → Intent classifier: trains model (evidence: `README.md`)
- Trajectories → Model training: feeds training
- Neighbor data → Model training: feeds training
- LiDAR features → Model training: feeds training
- Normalization stats → Model training: feeds training
- Model evaluation → Prediction metrics: computes (evidence: `README.md`)
- Model evaluation → Evaluation results: records results
- GMM predictions → Model evaluation: evaluates predictions
