EeveeDark: Binary Low-Light Video Enhancement

• EeveeDark is a low-light video enhancement framework that fuses sensor-level Bayer RAW frames with high-temporal event streams using binary neural computation.
• It efficiently recovers clean, bright, and temporally coherent outputs while avoiding motion artifacts by leveraging cyclic shift operations and event-guided skip gates.
• The method achieves a favorable quality-efficiency balance, making it suitable for embedded systems and robotic platforms in extreme darkness.

EeveeDark is a low-light video enhancement framework that combines sensor-level Bayer RAW frames with event streams inside a largely binary neural network, targeting the regime in which extreme darkness, temporal coherence, and resource-constrained deployment must be addressed simultaneously (Eker et al., 7 Jul 2026). It is formulated for sequences of noisy, dark RAW video frames and seeks to recover clean, bright, temporally coherent outputs while retaining computational efficiency suitable for embedded or robotic platforms. The method is explicitly positioned around three coupled design choices: sensor-level fusion rather than processed RGB input, event-guided multimodal enhancement rather than image-only restoration, and binary neural computation rather than full-precision video restoration throughout most of the network (Eker et al., 7 Jul 2026).

1. Problem formulation and design rationale

EeveeDark targets low-light video enhancement in settings where extreme darkness causes low signal-to-noise ratio, low contrast, color degradation, and unstable temporal behavior (Eker et al., 7 Jul 2026). The problem is not treated as a framewise enhancement task alone. Because the input is video, the method also seeks to avoid flicker and exploit temporal redundancy without introducing motion artifacts.

The framework is motivated by two shortcomings that the paper attributes to existing event-guided low-light enhancement methods. First, they often use processed RGB inputs, which lose sensor-level information important for faithful restoration in severe darkness. Second, they often rely on expensive alignment or attention modules, such as optical flow, which the paper identifies as unsuitable for edge deployment (Eker et al., 7 Jul 2026). EeveeDark therefore adopts sensor-level fusion, starting from Bayer RAW and combining it with event data rather than enhancing already processed RGB imagery.

The two sensing modalities are treated as complementary. RAW frames provide spatial richness, photometric fidelity, and sensor-level color and tonal information. Event streams provide high temporal resolution, large dynamic range, and motion and brightness-change cues that remain informative in difficult lighting (Eker et al., 7 Jul 2026). The paper’s central claim is that this multimodal input can compensate for the quality degradation ordinarily induced by binarization, thereby enabling a lightweight model that remains practical on resource-constrained hardware.

The binary design is not presented as an isolated efficiency device. Rather, the method argues that a Binary Neural Network (BNN) can become viable for restoration if the input representation itself is sufficiently informative. This suggests that EeveeDark should be understood less as a conventional BNN adapted to low light than as a sensor-fusion architecture in which binary computation is made workable by RAW-domain processing and event guidance.

2. Input representation and overall pipeline

At a high level, EeveeDark takes a sequence of packed Bayer RAW frames and synchronized event streams converted to voxel grids between frames, and predicts enhanced RAW frames, which are then converted to RGB using a standard ISP for visualization and evaluation (Eker et al., 7 Jul 2026).

For the RAW input, the paper defines a Bayer sequence

$$\{\mathbf{I}_B^1,\mathbf{I}_B^2,\dots,\mathbf{I}_B^T\}, \qquad \mathbf{I}_B^i \in \mathbb{R}^{H\times W}.$$

Each $2\times2$ Bayer block is packed into four channels, yielding

$$\mathbf{I}_p^i \in \mathbb{R}^{\frac{H}{2}\times \frac{W}{2}\times 4},$$

and amplified using a scaling factor $r$, although the exact value of $r$ is not stated in the paper excerpt (Eker et al., 7 Jul 2026).

For events, each event is represented as

$e_i=(x_i,y_i,t_i,p_i),$

with polarity $p_i\in\{+1,-1\}$. Events are grouped between consecutive frame times into sets $G_k$, then converted into voxel grids

$$E_k \in \mathbb{R}^{\frac{H}{2}\times \frac{W}{2}\times B}$$

using the polarity-based linear accumulation method of Rebecq et al., with

$B=5$

temporal bins in all experiments (Eker et al., 7 Jul 2026).

The processing pipeline contains seven stages: preprocessing; modality-specific binary encoders; lightweight multimodal fusion; a shift encoder with recurrent embeddings; an Event-Guided Skip Gate (EGSG); a shift decoder with recurrent embeddings; and a residual restoration module that outputs enhanced RAW (Eker et al., 7 Jul 2026). The architecture is described as having five key components: preprocessing and separate encoding of Bayer RAW and events; efficient multimodal fusion; a shift encoder carrying temporal information and using EGSG; a shift decoder refining and aligning temporal features; and a final restoration module whose output is later passed through an ISP.

This structure is explicitly distinct from a flow-based alignment pipeline. Temporal propagation is handled by grouped spatial-temporal shift operations from ShiftNet rather than optical-flow estimation (Eker et al., 7 Jul 2026). A plausible implication is that the method’s efficiency claim depends not only on binarization, but also on replacing motion estimation with shift-based temporal exchange.

3. Network architecture

EeveeDark uses two separate encoders, one for frames and one for events, and the paper states that they employ distribution-aware binary convolutions from BRVE (Eker et al., 7 Jul 2026). The first convolution in each encoder is kept full precision to preserve low-level detail from the RAW and event inputs. The encoder outputs are written as

$2\times2$0

The fusion block combines the two modalities by channel concatenation:

$2\times2$1

The fused representation is then computed through a binary $2\times2$2 branch and a full-precision $2\times2$3 branch:

$2\times2$4

The paper characterizes this as a deliberately lightweight fusion mechanism rather than an attention-based cross-modal transformer (Eker et al., 7 Jul 2026).

Temporal propagation begins in a shift encoder with recurrent embeddings:

$2\times2$5

followed by

$2\times2$6

This design propagates temporal information using cyclic channel shift rather than explicit motion compensation (Eker et al., 7 Jul 2026).

A principal novelty is the Event-Guided Skip Gate (EGSG), which uses event features to generate a gating map for the shifted encoder features. The event feature map is resized and projected:

$2\times2$7

then summarized through channel-wise statistics:

$2\times2$8

The gate is obtained as

$2\times2$9

and applied multiplicatively:

$$\mathbf{I}_p^i \in \mathbb{R}^{\frac{H}{2}\times \frac{W}{2}\times 4},$$0

The interpretation given in the paper is that event-derived statistics indicate where brightness changes and motion occur, allowing the gate to emphasize dynamic regions and suppress redundant ones (Eker et al., 7 Jul 2026).

The decoder operates on a local temporal window $$\mathbf{I}_p^i \in \mathbb{R}^{\frac{H}{2}\times \frac{W}{2}\times 4},$$1. Each gated feature map is split into stationary and shiftable components:

$$\mathbf{I}_p^i \in \mathbb{R}^{\frac{H}{2}\times \frac{W}{2}\times 4},$$2

The shiftable parts are cyclically redistributed as

$$\mathbf{I}_p^i \in \mathbb{R}^{\frac{H}{2}\times \frac{W}{2}\times 4},$$3

and a spatial shift is applied using

$$\mathbf{I}_p^i \in \mathbb{R}^{\frac{H}{2}\times \frac{W}{2}\times 4},$$4

with

$$\mathbf{I}_p^i \in \mathbb{R}^{\frac{H}{2}\times \frac{W}{2}\times 4},$$5

The decoded feature is then

$$\mathbf{I}_p^i \in \mathbb{R}^{\frac{H}{2}\times \frac{W}{2}\times 4},$$6

and the final restoration is residual:

$$\mathbf{I}_p^i \in \mathbb{R}^{\frac{H}{2}\times \frac{W}{2}\times 4},$$7

The output layer $$\mathbf{I}_p^i \in \mathbb{R}^{\frac{H}{2}\times \frac{W}{2}\times 4},$$8 is full precision (Eker et al., 7 Jul 2026).

The paper does not provide explicit quantization equations, straight-through estimator formulas, exact channel widths, or layer counts. It states only that most internal convolutions use binary weights and activations, while the first layer in each encoder, selected $$\mathbf{I}_p^i \in \mathbb{R}^{\frac{H}{2}\times \frac{W}{2}\times 4},$$9 fusion and gating layers, and the final output layer remain full precision (Eker et al., 7 Jul 2026).

4. Training objective and implementation protocol

Training is performed entirely in the RAW domain using only the Charbonnier loss:

$r$0

Here $r$1 denotes the predicted frame and $r$2 the ground-truth RAW frame (Eker et al., 7 Jul 2026). The paper explicitly does not include perceptual loss, SSIM loss, temporal consistency loss, adversarial loss, or an event-specific auxiliary loss.

Implementation details are reported as follows: PyTorch is used for training on an NVIDIA RTX A5000; the training unit is $r$3 patches from 10-frame sequences; augmentations are random cropping, horizontal flipping, vertical flipping; optimization uses Adam with

$r$4

training runs for 100,000 iterations; and the learning rate follows cosine annealing restart, decaying from

$r$5

The batch size is not stated in the excerpt (Eker et al., 7 Jul 2026).

The efficiency analysis is expressed through explicit complexity formulas:

$r$6

and

$r$7

Average per-frame FLOPs are reported on a 100-frame video at $r$8, while latency is estimated using daBNN on ARM64 mapping and energy through a 7nm device model following prior BNN work (Eker et al., 7 Jul 2026). The paper notes that direct GPU runtime is not meaningful because current GPU frameworks do not natively support binarized layers.

A methodological point follows from these choices. Because the loss remains purely photometric in the RAW domain despite the use of events, the event stream functions as an internal guidance modality rather than a source of explicit supervisory signal. This distinguishes EeveeDark from event-guided radiance-field methods such as Dark-EvGS, where event supervision appears directly in the optimization objective (Wu et al., 16 Jul 2025).

5. Datasets, evaluation settings, and empirical results

The experiments cover LLRVD, HUE, SDE, SDSD, and CEAR (Eker et al., 7 Jul 2026). The paper states that methods on LLRVD and HUE were retrained from scratch using public implementations. HUE has no reference ground truth for enhancement, so no-reference IQA metrics are used there. CEAR frames are warped into the event camera frame using calibrated intrinsics and extrinsics for downstream evaluation (Eker et al., 7 Jul 2026).

The comparison settings are divided into three groups. For RAW-to-RAW on LLRVD, baselines are BBCU, BRVE, ShiftNet, and FloRNN. For RGB-domain evaluation via ISP on LLRVD/HUE, the comparisons include those methods plus EvLight. For RGB benchmarks on SDE/SDSD, the paper additionally compares against SNR-Net, Uformer, Retinexformer, ELIE, eSL-Net, Liu et al., EvLight, E2VID+, and BRVE (Eker et al., 7 Jul 2026).

On LLRVD in the RAW domain, Table 1 reports:

Method	PSNR	SSIM	ST-RRED
FloRNN	37.20	0.960	0.042
ShiftNet	37.85	0.967	0.031
BBCU	36.52	0.950	0.058
BRVE	37.07	0.958	0.046
EeveeDark	37.51	0.962	0.039

These numbers show that EeveeDark improves over the prior binary methods BBCU and BRVE, with a reported gain of +0.99 dB PSNR over BBCU and +0.44 dB PSNR over BRVE, while approaching the quality of the full-precision models (Eker et al., 7 Jul 2026). The paper also reports that BRVE exhibits temporal flickering and that EeveeDark improves temporal stability.

After ISP on LLRVD, the paper reports 30.37 / 0.850 / 0.189 for EeveeDark, compared with 29.58 / 0.821 / 0.233 for BRVE, 28.72 / 0.795 / 0.373 for BBCU, 30.28 / 0.845 / 0.185 for EvLight, 30.53 / 0.853 / 0.176 for FloRNN, and 32.01 / 0.889 / 0.095 for ShiftNet (Eker et al., 7 Jul 2026). On HUE, EeveeDark yields 0.177 CLIP-IQA, 0.158 MANIQA, 0.605 TOPIQ-NR, and 5.358 NIQE, with NIQE being the best among the listed methods (Eker et al., 7 Jul 2026).

On SDE, EeveeDark reports 21.99 PSNR, 24.32 PSNR*, and 0.707 SSIM, compared with 21.52 / 23.46 / 0.672 for BRVE and 22.83 / 25.21 / 0.760 for EvLight (Eker et al., 7 Jul 2026). On SDSD, it reports 26.66 / 28.69 / 0.859, compared with 24.14 / 27.72 / 0.816 for BRVE and 27.60 / 30.02 / 0.875 for EvLight (Eker et al., 7 Jul 2026). The paper’s qualitative interpretation is that BRVE oversmooths and amplifies noise in dark regions, whereas EeveeDark restores finer structures and local contrast using event information.

The ablation results isolate two central components. The full EeveeDark (RAW + Events) model obtains 37.51 / 0.962 / 0.039 on LLRVD; w/o Event Encoder (RAW only) drops to 37.07 / 0.958 / 0.046; and w/o EGSG yields 37.44 / 0.962 / 0.041 (Eker et al., 7 Jul 2026). The paper also compares enhancement settings: RAW2RAW + ISP gives 30.37 / 0.8501 / 0.2021, RAW2RGB gives 28.26 / 0.8492 / 0.2393, and RGB2RGB gives 26.79 / 0.8165 / 0.4328, which the authors use to support the claim that operation in the RAW domain matters (Eker et al., 7 Jul 2026).

6. Efficiency, downstream utility, and limitations

Efficiency is a primary part of EeveeDark’s identity. The model is reported at 1.66G FLOPs, 0.35M parameters, ~588 ms, and 0.78 mJ, compared with 1.49G / 0.30M / ~528 ms / 0.70 mJ for BRVE, 1.47G / 0.30M / ~521 ms / 0.69 mJ for BBCU, 24.57G / 10.49M / ~7709 ms / 20.76 mJ for FloRNN, 32.87G / 13.38M / ~10313 ms / 27.78 mJ for ShiftNet, and 48.54G / 22.73M / ~15227 ms / 32.57 mJ for EvLight (Eker et al., 7 Jul 2026). The empirical trade-off claimed by the paper is therefore not maximal quality at any cost, but a favorable quality-efficiency balance relative to both binary baselines and full-precision alternatives.

The paper goes beyond image enhancement metrics and evaluates downstream robotic perception. On object detection on SDE using YOLOv11-s, mAP@0.5 improves from 0.21 on low-light input to 0.47 with BRVE and 0.73 with EeveeDark (Eker et al., 7 Jul 2026). For monocular depth estimation, the paper reports on SDE: low light AbsRel 1.390, RMSE 1.705, $r$9, $r$0, $r$1; BRVE 0.921, 1.101, 0.523, 0.785, 0.869; EeveeDark 0.810, 0.983, 0.568, 0.810, 0.893. On CEAR, low light gives 0.912, 1.860, 0.191, 0.415, 0.585; BRVE 0.460, 1.230, 0.353, 0.638, 0.812; and EeveeDark 0.414, 1.060, 0.381, 0.685, 0.863 (Eker et al., 7 Jul 2026). For visual SLAM on CEAR, ATE RMSE improves across nearly all sequences; examples given are around_bldg: 39.93 → 36.36 → 30.56, parking_lot1: 40.46 → 11.50 → 6.11, and mocap1: 1.77 → 1.50 → 0.27 for low light, BRVE, and EeveeDark respectively (Eker et al., 7 Jul 2026).

The paper explicitly presents EeveeDark as the first end-to-end trainable BNN for low-light video enhancement that jointly processes asynchronous events and sensor-level RAW video (Eker et al., 7 Jul 2026). In the broader low-light event-vision landscape, this places it in a different niche from Dark-EvGS, which focuses on static-scene radiance field reconstruction and bright novel-view synthesis from dark event and frame data rather than video enhancement (Wu et al., 16 Jul 2025). EeveeDark is therefore a restoration framework rather than a radiance-field method.

The limitations acknowledged in the paper are tied directly to the modality design. When motion is minimal, events become sparse and the event guidance weakens. Under extremely low photon counts, RAW itself is very noisy, which can degrade enhancement quality, produce residual noise, and harm color fidelity (Eker et al., 7 Jul 2026). The system is therefore less effective in static, ultra-dark scenes where both modalities are weak. The most natural deployment settings identified by the paper are edge devices, embedded robotic platforms, mobile or battery-constrained vision systems, and autonomous navigation in low-light environments (Eker et al., 7 Jul 2026).

Taken together, EeveeDark is best understood as a multimodal, RAW-domain, efficiency-oriented video restoration framework in which event guidance is used not for explicit event reconstruction, but for temporal modulation and feature refinement inside a predominantly binary architecture. Its contribution lies in the joint integration of sensor-level RAW processing, event-guided low-light video enhancement, and binary neural computation within a single end-to-end trainable system (Eker et al., 7 Jul 2026).