---
title: 'EfficientDet-D7: Scalable Object Detector'
url: https://www.emergentmind.com/topics/efficientdet-d7
type: topic
---

# EfficientDet-D7: Scalable Object Detector

EfficientDet-D7 is a state-of-the-art one-stage object detector designed for efficient and accurate multi-scale object detection. Emerging from the EfficientDet family, it integrates a weighted bi-directional feature pyramid network (BiFPN), compound scaling, and an EfficientNet-B6 backbone for extracting image features. EfficientDet-D7 demonstrates a favorable trade-off between accuracy and computational cost, achieving 55.1% AP on COCO test-dev with a model containing 77 million parameters and 410 billion FLOPs when using single model and single scale, while being significantly smaller and faster than preceding systems [1911.09070].

## 1. Architecture Overview

EfficientDet-D7 comprises three principal components:

1. **Backbone:** EfficientNet-B6, pretrained on ImageNet, serves as the backbone and is responsible for encoding multi-scale features from the input image.
2. **BiFPN:** The Bidirectional Feature Pyramid Network is a sequence of identical layers that fuses features from backbone stages $P_3$ through $P_7$ using both top-down and bottom-up paths.
3. **Class/Box Prediction Heads:** Lightweight subnetworks are attached at each pyramid level and are responsible for classifying objects and regressing their bounding boxes.

This architectural composition realizes a high degree of parameter efficiency and enables robust multi-scale object representation.

## 2. BiFPN: Bidirectional Feature Pyramid Network

BiFPN is the critical component for cross-scale feature fusion. In EfficientDet-D7, five input features ($P_3$ to $P_7$) from the backbone, corresponding to scales 1/8, 1/16, 1/32, 1/64, and 1/128 of the $1536 \times 1536$ input, are utilized (for $P_3$, the resolution is $192 \times 192$; for $P_7$, $12 \times 12$).

Eight stacked BiFPN layers are employed. Each layer incorporates both top-down and bottom-up fusion, augmented by shortcut connections that maximize information flow while omitting nodes that aggregate a single input. 

Feature fusion within BiFPN is realized via fast normalized fusion. For $N$ input features $\{I_j\}_{j=1}^N$ at level $i$, the output is

\[
P_i = \operatorname{Conv}\left(\frac{\sum_{j=1}^N w_j \, I_j}{\sum_{j=1}^N w_j + \epsilon}\right),
\]

where each $w_j \geq 0$ is a learnable weight (with non-negativity enforced via ReLU), and $\epsilon = 10^{-4}$ prevents division by zero. This approach enables the network to learn the relative contribution of each input feature map with minimal overhead.

## 3. Compound Scaling Method

EfficientDet implements a compound scaling strategy to jointly scale network width, depth, and resolution using a global coefficient $\phi$. Rather than disjointly tuning these dimensions, EfficientDet employs heuristic-based rules:

- **Backbone:** EfficientDet-D7 uses EfficientNet-B6, scaling width and depth following the EfficientNet series.
- **BiFPN Width/Depth:** 
  - Width: $W_{\mathrm{bifpn}} = 64 \times 1.35^{\phi}$ 
  - Depth: $D_{\mathrm{bifpn}} = 3 + \phi$
- **Prediction Head Depth:** $D_{\mathrm{heads}} = 3 + \lfloor \phi / 3 \rfloor$
- **Input Resolution:** $R_{\mathrm{in}} = 512 + 128\,\phi$

For EfficientDet-D7 ($\phi = 7$), this yields a $1536\times1536$ input, eight BiFPN layers of approximately 384 channels each, and five backbone stages from EfficientNet-B6.

## 4. Model Complexity and Empirical Results

EfficientDet-D7x, an extended version of D7 substituting EfficientNet-B7 as the backbone, demonstrates:

| Model | Parameters (M) | FLOPs (B) | COCO test-dev AP (%) |
|-------|----------------|-----------|----------------------|
| D7x   | 77             | 410       | 55.1                 |

Detailed COCO test-dev performance for D7x:

- $AP_{50}$: 74.3%
- $AP_{75}$: 59.9%
- $AP_S \approx 38\%$
- $AP_M \approx 60\%$
- $AP_L \approx 70\%$

Relative to AmoebaNet + NAS-FPN (+AutoAugment), EfficientDet-D7x attains a $+4.4\%$ $AP$ improvement, with $2.4\times$ fewer parameters (77M vs. 185M) and $3.2\times$ fewer FLOPs (410B vs. 1317B).

## 5. Training and Implementation Details

Key training and implementation protocols essential for achieving EfficientDet-D7 performance are as follows:

- **Optimization:** SGD with momentum 0.9 and weight decay $4 \times 10^{-5}$.
- **Learning Rate:** Linear warmup from 0 to 0.16 in the first epoch, then cosine decay to zero.
- **Batch Size:** 128, distributed across 32 TPUv3 cores; training is conducted for 600 epochs for D7/D7x.
- **Data Augmentation:** Horizontal flip, scale jittering in $[0.1, 2.0]$ range, followed by random cropping to $1536 \times 1536$.
- **Loss Function:** Focal loss (parameters: $\alpha = 0.25$, $\gamma = 1.5$), anchor aspect ratios $\{1/2, 1, 2\}$.
- **Normalization and Activation:** Synchronized BatchNorm (decay 0.99, $\epsilon = 10^{-3}$) after every convolution, SiLU (Swish-1) activations, Exponential Moving Average (EMA) of weights (decay 0.9998).
- **Inference:** Single-scale prediction with soft-NMS postprocessing.

## 6. Significance and Comparative Analysis

EfficientDet-D7 demonstrates that state-of-the-art object detection performance on COCO test-dev can be achieved with substantially reduced parameter count and arithmetic operations relative to prior detectors. Its architecture, combining EfficientNet-B6, depthwise-separable convolutions in BiFPN, weighted multi-scale fusion, and compound scaling, enables a new benchmark for single-model single-scale detection accuracy and complexity.

The model’s modular design permits adaptation to a broad range of deployment scenarios with varying computational constraints, as evidenced by scaling approaches that enable the entire EfficientDet family to fit diverse efficiency-accuracy trade-off envelopes [1911.09070]. A plausible implication is that further advances in backbone architectures or feature fusion strategies might yield additional improvements under the same compound scaling regime.

Source: https://www.emergentmind.com/topics/efficientdet-d7