AI Preprocessing Difficulty — Text, Image, Audio, Video, 3D in Order of Complexity

All AI inference follows this structure:

processed_input = preprocess(raw_data)   # ← differs by domain
output = model(processed_input)          # ← always the same
result = postprocess(output)             # ← differs by domain

model(input) just takes a PyTorch tensor and returns a tensor. The challenge is converting raw data into that tensor.

Level 1: Text — Tokenization (★☆☆☆☆)

Easiest. Start here.

tokens = tokenizer("Hello world")["input_ids"]
# → [15496, 995]  — lookup table, that's it

Why easy: 1D sequence, discrete integers, deterministic, tiny data.

Level 2: Tabular Data — Normalize (★★☆☆☆)

features = (features - features.mean()) / features.std()

Already numbers. Just scale them.

Level 3: Image — Resize + Normalize (★★☆☆☆)

preprocess = transforms.Compose([Resize(224), CenterCrop(224), ToTensor(), Normalize(...)])
tensor = preprocess(Image.open("cat.jpg"))  # [3, 224, 224]

Why harder than text: continuous values, larger data, need normalization stats.

Level 4: Audio — STFT + Mel (★★★★☆)

mel_spec = MelSpectrogram(sr=44100, n_fft=2048, n_mels=80)(wav)
log_mel = torch.log(mel_spec + 1e-9)  # [80, time_frames]

Why hard: need Fourier transform understanding, model-specific hyperparameters, non-intuitive frequency domain.

Level 5: Video — Frames + Temporal (★★★★☆)

frames = video[sampled_indices]  # [16, 3, 224, 224]

Image preprocessing × N frames + temporal dimension + sampling strategy.

Level 6: 3D / Point Clouds (★★★★★)

points -= centroid; points /= max_dist  # normalize
points = farthest_point_sampling(points, 1024)  # [1024, 3]

Hardest: irregular data, order-invariant, coordinate alignment, immature ecosystem.

Study Order

1. Text → 2. Image → 3. Audio → 4. Video → 5. 3D

Text has the shortest distance to model(input). That's why it's the best starting point.